CN112486875A - Multi-chip package link - Google Patents
Multi-chip package link Download PDFInfo
- Publication number
- CN112486875A CN112486875A CN202011344684.9A CN202011344684A CN112486875A CN 112486875 A CN112486875 A CN 112486875A CN 202011344684 A CN202011344684 A CN 202011344684A CN 112486875 A CN112486875 A CN 112486875A
- Authority
- CN
- China
- Prior art keywords
- data
- lanes
- logic
- hard
- protocol
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/38—Information transfer, e.g. on bus
- G06F13/40—Bus structure
- G06F13/4004—Coupling between buses
- G06F13/4022—Coupling between buses using switching circuits, e.g. switching matrix, connection or expansion network
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/14—Handling requests for interconnection or transfer
- G06F13/36—Handling requests for interconnection or transfer for access to common bus or bus system
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/38—Information transfer, e.g. on bus
- G06F13/40—Bus structure
- G06F13/4063—Device-to-bus coupling
- G06F13/4068—Electrical coupling
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- Mathematical Physics (AREA)
- Information Transfer Systems (AREA)
Abstract
A system on chip such as a logical PHY may be partitioned into hard IP blocks with fixed routing and soft IP blocks with flexible routing. Each hard IP block may provide a fixed number of lanes. P hard IP blocks are used, where each block provides n data lanes, and all h n p hard IP data lanes are provided. Wherein the system design requires all k data lanes, possibly k ≠ h, such that [ k/n ] hard IP block provides h ═ n × p available hard IP data lanes. In this case, h-k lanes may be disabled. In the case of path reversal, e.g., between hard and soft IPs, bowtie routing may be avoided through the use of multiplexer-like programmable switches within the soft IP.
Description
This application is a divisional application entitled 'multi-chip package link' with application number 201680012402.0 filed on 22/2/2016.
Cross Reference to Related Applications
This application claims benefit and priority from U.S. non-provisional patent application No.14/669,975 entitled "multi hip PACKAGE LINK" filed 3/26/2015, the entire contents of which are incorporated herein by reference.
Technical Field
The present disclosure relates to computing systems, and in particular (but not exclusively) to point-to-point interconnects.
Background
Advances in semiconductor processing and logic design have allowed an increase in the amount of logic that may be present on integrated circuit devices. As a corollary, computer system configurations have evolved from single or multiple integrated circuits in a system to multiple cores, multiple hardware threads, and multiple logical processors present on separate integrated circuits, as well as other interfaces integrated within such processors. A processor or integrated circuit typically includes a single physical processor die, where the processor die may include any number of cores, hardware threads, logical processors, interfaces, memory, controller hubs, and the like.
Smaller computing devices are becoming increasingly popular due to the greater ability to supply more processing power in smaller packages. Smart phones, tablets, ultra-thin laptops, and other user devices have grown exponentially. However, these smaller devices rely on servers for both data storage and complex processing beyond physical size. Thus, the demand for a high performance computing market (i.e., server space) is also increasing. For example, in modern servers, there is typically not only a single processor with multiple cores, but also multiple physical processors (also referred to as multiple sockets) for increasing computational power. But as processing power increases with the number of devices in a computing system, communication between the outlet and other devices becomes more critical.
In fact, interconnects have evolved from more traditional multi-drop buses, which deal primarily with electrical communications, to fully expanded interconnect architectures that facilitate rapid communications. Unfortunately, as the demand for future processors to consume at even higher rates increases, the corresponding demand is within the capabilities of existing interconnect architectures.
Drawings
FIG. 1 illustrates an embodiment of a computing system including an interconnect architecture.
Fig. 2 illustrates an embodiment of an interconnect architecture including a layered stack.
FIG. 3 illustrates an embodiment of a request or packet to be generated or received within an interconnect architecture.
Fig. 4 illustrates an embodiment of a transmitter and receiver pair for an interconnect architecture.
Fig. 5 illustrates an embodiment of a multi-chip package.
Fig. 6 is a simplified block diagram of a multi-chip package link (MCPL).
Fig. 7 is a representation of an exemplary signaling over an exemplary MCPL.
Fig. 8 is a simplified block diagram illustrating data paths in an exemplary MCPL.
Fig. 9 is a simplified block diagram illustrating an exemplary crosstalk cancellation technique in an embodiment of MCPL.
Fig. 10 is a simplified circuit diagram illustrating an exemplary crosstalk cancellation component in an embodiment of an MCPL.
Fig. 11 is a simplified block diagram of the MCPL.
Fig. 12 is a simplified block diagram of an MCPL interfacing with upper layer logic of multiple protocols using a Logical PHY Interface (LPIF).
Fig. 13 is a representation of exemplary signaling over an exemplary MCPL in connection with the recovery of a link.
Fig. 14A-14C are exemplary bit maps of data on the lanes of an exemplary MCPL.
FIG. 15 is a representation of a portion of an exemplary link state machine.
Fig. 16 is a representation of a flow associated with exemplary centering of a link.
Fig. 17 is a representation of an exemplary link state machine.
FIG. 18 is a representation of signaling for entering a low power state.
FIG. 19 illustrates an embodiment of a block diagram for a computing system including a multicore processor.
FIG. 20 illustrates another embodiment of a block diagram for a computing system including a multicore processor.
FIG. 21 illustrates an embodiment of a block diagram for a processor.
FIG. 22 illustrates another embodiment of a block diagram for a computing system including a processor.
FIG. 23 illustrates an embodiment of blocks for a computing system including multiple processors.
Fig. 24 illustrates an exemplary system implemented as a system on a chip (SoC).
Fig. 25 shows an example of signal transmission in the embodiment.
FIG. 26 is an exemplary flow chart of a method for encoding data for which an explicit STREAM path does not exist.
FIG. 27 is an example flow diagram of a method for encoding data for which no explicit STREAM path exists.
FIG. 28 is an exemplary flow chart of a method for performing centering of a reference voltage.
Fig. 29A and 29B are examples of crosstalk patterns.
Fig. 30 is an exemplary illustration of eye centering.
Fig. 31 is an exemplary flow chart of a method for eye centering.
Fig. 32 is an exemplary block diagram of a system including hard IP and soft IP.
Fig. 33 is an exemplary block diagram of an interface between a hard IP block and a soft IP block.
FIG. 34 is an example of a system for handling lane inversion for a programmable switch with a class multiplexer.
Like reference numbers and designations in the various drawings indicate like elements.
Detailed Description
In the following description, numerous specific details are set forth, such as examples of specific types of processor and system configurations, specific hardware structures, specific architectural and microarchitectural details, specific register configurations, specific instruction types, specific system components, specific measurements/heights, specific processor pipeline stages and operations, etc., in order to provide a thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that these specific details need not be employed to practice the present invention. In other instances, well known components or methods, such as specific and alternative processor architectures, specific logic circuits/code for described algorithms, specific firmware code, specific interconnect operation, specific logic configurations, specific manufacturing techniques and materials, specific compiler implementations, specific expression of algorithms in code, specific power down and gate control techniques/logic in computer systems, and other specific operational details of computer systems, have not been described in detail in order to avoid unnecessarily obscuring the present invention.
Although the following embodiments may be described with reference to energy conservation and energy efficiency in a particular integrated circuit, such as a computing platform or microprocessor, other embodiments are applicable to other types of integrated circuits and logic devices. Similar techniques and teachings of embodiments described herein may be applied to other types of circuits or semiconductor devices that may also benefit from better energy efficiency and energy conservation. For example, the disclosed embodiments are not limited toDesktop computer systems or UltrabooksTM. They may also be used in other devices, such as handheld devices, tablet computers, other thin notebook computers, system on a chip (SOC) devices, and embedded applications. Some examples of handheld devices include cellular telephones, internet protocol devices, digital cameras, Personal Digital Assistants (PDAs), and handheld PCs. Embedded applications typically include microcontrollers, Digital Signal Processors (DSPs), systems on a chip, network computers (netpcs), set-top boxes, hubs, Wide Area Network (WAN) switches, or any other system that can perform the functions and operations taught below. Furthermore, the apparatus, methods, and systems described herein are not limited to physical computing devices, but may also relate to software optimization for energy conservation and efficiency. As will become apparent in the following description, embodiments of the methods, apparatus and systems described herein (whether in reference to hardware, firmware, software or a combination thereof) are critical to the "green technology" future in balance with performance considerations.
As computing systems have advanced, the components therein have become more complex. Accordingly, interconnect architectures that couple and communicate between components are also increasing in complexity to ensure that bandwidth requirements for optimal component operation are met. Furthermore, different market areas require different aspects of the interconnect architecture to accommodate market demands. For example, servers require higher performance, and mobile ecosystems can sometimes sacrifice overall performance to save power. However, the single purpose of most architectures is to provide the highest possible performance with the greatest power savings. Many interconnects are discussed below that would likely benefit from aspects of the invention described herein.
An interconnect fabric architecture includes a Peripheral Component Interconnect (PCI) express (PCIe) architecture. The primary goal of PCIe is to enable components and devices from different vendors to interoperate in an open architecture and span multiple market domains; clients (desktop and mobile), servers (standard and enterprise), and embedded and communication devices. PCI Express (peripheral component interconnect Express) is a high performance, general purpose I/O interconnect defined for a variety of future computing and communication platforms. Some PCI attributes (e.g., its usage model, load-store architecture, and software interface) have been maintained through its modifications, while previous parallel bus implementations have been replaced by highly scalable full serial interfaces. Newer versions of PCI Express utilize point-to-point interconnects, switch-based technologies, and encapsulated protocols to deliver new levels of performance and functionality. Power management, quality of service (QoS), hot-plug/hot-plug support, data integrity and error handling are among some of the advanced functions supported by PCI Express.
Referring to fig. 1, an embodiment of a fabric including point-to-point links interconnecting a set of components is shown. System 100 includes a processor 105 and a system memory 110 coupled to a controller hub 115. Processor 105 includes any processing element such as a microprocessor, host processor, embedded processor, co-processor, or other processor. The processor 105 is coupled to the controller hub 115 through a Front Side Bus (FSB) 106. In one embodiment, the FSB 106 is a serial point-to-point interconnect as described below. In another embodiment, the link 106 includes a serial, differential interconnect architecture that conforms to different interconnect standards.
The system memory 110 includes any memory device (e.g., Random Access Memory (RAM), non-volatile (NV) memory, or other memory accessible by devices in the system 100). System memory 110 is coupled to controller hub 115 through memory interface 116. Examples of memory interfaces include Double Data Rate (DDR) memory interfaces, double channel DDR memory interfaces, and dynamic ram (dram) memory interfaces.
In one embodiment, controller hub 115 is a root hub, root complex, or root controller in a peripheral component interconnect express (PCIe or PCIe) interconnect hierarchy. Examples of controller hub 115 include a chipset, a Memory Controller Hub (MCH), a northbridge, an Interconnect Controller Hub (ICH), a southbridge, and a root controller/hub. In general, the term chipset refers to two physically separate controller hubs, i.e., a Memory Controller Hub (MCH) coupled to an Interconnect Controller Hub (ICH). It should be noted that current systems typically include an MCH integrated with the processor 105, while the controller 115 is used to communicate with I/O devices in a similar manner as described below. In some embodiments, peer-to-peer routing is optionally supported by root complex 115.
Here, controller hub 115 is coupled to switch/bridge 120 through serial link 119. Input/output modules 117 and 121 (which may also be referred to as interfaces/ports 117 and 121) include/implement a layered protocol stack to provide communication between controller hub 115 and switch 120. In one embodiment, multiple devices can be coupled to the switch 120.
Switch/bridge 120 routes packets/messages from device 125 upstream (i.e., toward the level of the root complex) to controller hub 115 and downstream (i.e., away from the level of the root controller down) from processor 105 or system memory 110 to device 125. In one embodiment, the switch 120 is referred to as a logical component of a plurality of virtual PCI-to-PCI bridge devices. Device 125 includes any internal or external device or component to be coupled to an electronic system, such as I/O devices, Network Interface Controllers (NICs), add-in cards, audio processors, network processors, hard drives, storage devices, CD/DVD ROMs, monitors, printers, mice, keyboards, routers, portable storage devices, firewire devices, Universal Serial Bus (USB) devices, scanners, and other input/output devices. Generally, in PCIe terminology, for example, devices are referred to as endpoints. Although not specifically shown, device 125 may include a PCIe to PCI/PCI-X bridge to support legacy or other versions of PCI devices. Endpoint devices in PCIe are typically classified as root complexes of legacy, PCIe, or integrated endpoints.
Turning to fig. 2, an embodiment of a layered protocol stack is shown. Layered protocol stack 200 includes any form of layered communication stack, such as a Quick Path Interconnect (QPI) stack, a PCie stack, a next generation high performance computing interconnect stack, or other layered stack. Although the discussion below with reference to fig. 1-4 is related to a PCIe stack, the same concepts may be applied to other interconnect stacks. In one embodiment, protocol stack 200 is a PCIe protocol stack that includes a transaction layer 205, a link layer 210, and a physical layer 220. Interfaces such as interfaces 117, 118, 121, 122, 126, and 131 in fig. 1 may be represented as a communication protocol stack 200. A representation as a communication protocol stack may also be referred to as a module or interface implementing/including the protocol stack.
PCI Express uses packets to transfer information between components. Packets are formed in the transaction layer 205 and the data link layer 210 to convey information from the transmitting component to the receiving component. As transmitted packets flow through other layers, they are extended with additional information needed for packets processed in those layers. On the receiving side, reverse processing occurs and the packet is transformed from its physical layer 220 representation to a data link layer 210 representation and finally (transaction layer packet) into a form that can be processed by the transaction layer 205 of the receiving device.
Transaction layer
In one embodiment, transaction layer 205 is used to provide an interface between the processing cores of the device and the interconnect architecture (such as data link layer 210 and physical layer 220). In this regard, the primary responsibility of the transaction layer 205 is the assembly and disassembly of packets (i.e., transaction layer packets or TLPs). Transaction layer 205 typically manages credit-based flow control for TLPs. PCIe implementations split transactions, i.e., transactions with requests and responses separated by time, allowing the link to carry other traffic while the target device collects the data of the responses.
Furthermore, PCIe utilizes credit-based flow control. In this scheme, the device notifies each of the receive buffers in the transaction layer 205 of the initial amount of credits. An external device (e.g., controller hub 115 in figure 1) at the opposite end of the link counts the number of credits consumed by each TLP. If the transaction does not exceed the credit limit, the transaction may be transferred. When a response is received, the total number of credits is restored. An advantage of the credit scheme is that the delay in credit return does not affect performance provided that no credit limit is encountered.
In one embodiment, the four transaction address spaces include a configuration address space, a memory address space, an input/output address space, and a message address space. The memory space transactions include one or more of read requests and write requests for transferring data to/from memory-mapped locations. In one embodiment, memory space transactions can use two different address formats, for example, a short address format such as a 32-bit address, or a long address format such as a 64-bit address. The configuration space transaction is used to access a configuration space of the PCIe device. Transactions to the configuration space include read requests and write requests. Message space transactions (or simply messages) are defined to support in-band communication between PCIe agents.
Thus, in one embodiment, the transaction layer 205 assembles the packet header/payload 206. The format for the current packet header/payload may be found in the PCIe specification of the PCIe specification website.
Referring quickly to FIG. 3, an embodiment of a PCIe transaction descriptor is shown. In one embodiment, transaction descriptor 300 is a mechanism for conveying transaction information. In this regard, the transaction descriptor 300 supports the identification of transactions in the system. Other potential uses include tracking modifications to default transaction ordering and association of transactions with channels.
According to one embodiment, the local transaction identifier field 308 is a field generated by the requesting agent and is unique to all pending requests that require completion by the requesting agent. Further, in this example, source identifier 310 uniquely identifies the requester agent within the PCIe hierarchy. Thus, along with the source ID 310, the local transaction identifier 308 field provides global identification of transactions within the hierarchy domain.
The attributes field 304 specifies the nature and relationship of the transaction. In this regard, the attribute field 304 may be used to provide additional information that allows modification of the default handling of the transaction. In one embodiment, attribute fields 304 include a priority field 312, a reserved field 314, an ordering field 316, and a no snoop field 318. Here, the priority subfield 312 may be modified by the initiator to assign a priority to the transaction. The reserved attributes field 314 is reserved for future or vendor defined use. Possible usage models using priority or security attributes may be implemented using reserved attribute fields.
In this example, the sort attributes field 316 is used to provide optional information conveying the sort type that the default sort rule may be modified. According to one exemplary embodiment, the ordering attribute of "0" indicates a default ordering rule for the application, wherein the ordering attribute of "1" indicates relaxed ordering, wherein writes may be passed in the same direction and read completions may be passed in the same direction. Snoop attributes field 318 is used to determine whether a transaction is snooped. As shown, channel ID field 306 identifies the channel with which the transaction is associated.
Link layer
Link layer 210 (also referred to as data link layer 210) acts as an intermediate stage between transaction layer 205 and physical layer 220. In one embodiment, it is the responsibility of the data link layer 210 to provide a reliable mechanism for exchanging Transaction Layer Packets (TLPs) between two components in a link. One side of the data link layer 210 accepts a TLP assembled by the transaction layer 205, applies the packet sequence identifier 211 (i.e., identification number or packet number), calculates and applies an error detection code (i.e., CRC 212), and submits the modified TLP to the physical layer 220 for transmission across the physical layer to an external device.
Physical layer
In one embodiment, physical layer 220 includes a logical sub-block 221 and an electrical sub-block 222 to physically transmit packets to an external device. Here, the logical sub-block 221 is responsible for the "digital" function of the physical layer 221. In this regard, the logical sub-block includes a transmit section that prepares outgoing information sent by the physical sub-block 222, and a receiver section that identifies and prepares received information before passing it to the link layer 210.
The physical block 222 includes a transmitter and a receiver. The transmitter is provided by a logical sub-block 221 with symbols, which the transmitter serializes and transmits to an external device. The receiver is supplied with serialized symbols from an external device, and transforms the received signal into a bit stream. The bit stream is deserialized and supplied to the logical sub-block 221. In one embodiment, an 8b/10b transmission code is employed, wherein ten bit symbols are transmitted/received. Here, special symbols are used to frame the packets with frame 223. Additionally, in one example, the receiver also provides a symbol clock recovered from the incoming serial stream.
As described above, although transaction layer 205, link layer 210, and physical layer 220 are described with reference to a particular embodiment of a PCIe protocol stack, the layered protocol stack is not so limited. Indeed, any layered protocol may be included/implemented. As an example, a port/interface represented as a layered protocol includes: (1) a first layer for assembling packets, i.e., a transaction layer; a second layer, i.e., a link layer, that serializes packets; and a third layer, i.e., a physical layer, for transmitting packets. As a specific example, a Common Standard Interface (CSI) layered protocol is utilized.
Referring next to FIG. 4, an embodiment of a PCIe serial point-to-point fabric is shown. Although an embodiment of a PCIe serial point-to-point link is shown, the serial point-to-point link is not so limited as it includes any transmission path for communicating serial data. In the illustrated embodiment, the basic PCIe link includes two low voltage differential drive signal pairs: a transmit pair 406/411 and a receive pair 412/407. Thus, device 405 includes transmit logic 406 for transmitting data to device 410 and receive logic 407 for receiving data from device 410. In other words, two transmit paths (i.e., paths 416 and 417), and two receive paths (i.e., paths 418 and 419) are included in the PCIe link.
A transmission path refers to any path for transmitting data, such as a transmission line, a copper line, an optical line, a wireless communication channel, an infrared communication link, or other communication path. The connection between two devices (e.g., device 405 and device 410) is referred to as a link (e.g., link 415). A link may support one lane — each lane represents a set of differential signal pairs (one pair for transmission and one pair for reception). To scale bandwidth, a link may aggregate multiple lanes indicated by xN, where N is any supported link width, e.g., 1, 2, 4, 8, 12, 16, 32, 64, or wider.
A differential pair refers to two transmission paths (e.g., lines 416 and 417) for transmitting differential signals. As an example, when line 416 switches from a low voltage level to a high voltage level (i.e., a rising edge), line 417 drives from a high logic level to a low logic level (i.e., a falling edge). Differential signals may show better electrical characteristics, such as better signal integrity (i.e., cross-coupling, voltage overshoot/undershoot, ringing, etc.). This allows for a better timing window, which enables faster transmission frequencies.
Fig. 5 is a simplified block diagram 500 illustrating an exemplary multi-chip package 505, the multi-chip package 505 including two or more chips, or dies (e.g., 510, 515) communicatively connected using an exemplary multi-chip package link (MCPL) 520. Although fig. 5 shows an example of two (or more) dies that are interconnected using an example MCPL520, it should be understood that the principles and features described herein with respect to embodiments of MCPLs may be applied to any interconnect or link that connects a die (e.g., 510) and other components, including embodiments that connect two or more dies (e.g., 510, 515), a die (or chip) to another component outside of a die, a die (e.g., 505) outside of another device or package, a die to a BGA package, a Patch On Interposer (POINT), among potentially other examples.
In general, a multi-chip package (e.g., 505) may be an electronic package in which a plurality of Integrated Circuits (ICs), semiconductor dies, or other discrete components (e.g., 510, 515) are packaged onto a unified substrate (e.g., silicon or other semiconductor substrate) to facilitate use of the components as a combination of individual components (e.g., as with a larger IC). In some instances, the larger component (e.g., die 510, 515) may itself be an IC system, such as a system on a chip (SoC) that includes multiple components (e.g., 525 and 540 and 545) on a device (e.g., on a single die (e.g., 510, 515)), a multi-processor chip, or other component. The multi-chip package 505 may provide flexibility for building complex and varied systems from possibly multiple discrete components and systems. For example, each of the dies 510, 515 may be fabricated utilizing the silicon substrate of the package 505 provided by a third entity or otherwise provided by two different entities, among many other examples. Further, the die and other components within the multi-chip package 505 may themselves include interconnects or other communication structures (e.g., 535, 550) that provide an infrastructure for communication between components (e.g., 525-530 and 540-545, respectively) within the devices (e.g., 510, 515). The various components and interconnects (e.g., 535, 550) may potentially support or use a number of different protocols. Further, communications between the dies (e.g., 510, 515) may potentially include transactions between various components on the dies over a number of different protocols. Designing mechanisms to provide communication between chips (or dies) on a multi-chip package can be challenging, with traditional solutions employing highly specialized, expensive, and package-specific solutions based on the particular combination of components (and required transactions) that are attempting to be interconnected.
Examples, systems, algorithms, devices, logic, and features described within this specification may address at least some of the issues identified above, including potentially many other issues not explicitly mentioned herein. For example, in some embodiments, a high bandwidth, low power, low latency interface may be provided to connect a host device (e.g., CPU) or other device to a companion chip within the same package as the host. Such multi-chip package links (MCPLs) may support multiple package options, multiple I/O protocols, and reliability, availability, and serviceability (RAS) features. Further, the physical layer (PHY) may include electrical and logical layers, and may support longer channel lengths, including channel lengths up to about 45mm and in some cases exceeding about 45 mm. In some embodiments, the exemplary MCPL may operate at high data rates, including data rates in excess of 8-10 Gb/sec.
In one exemplary embodiment of the MCPL, the PHY electrical layer may improve traditional multi-channel interconnect solutions (e.g., multi-channel DRAM I/O), extend data rates and channel configurations, for example, by a number of features including, among potentially other examples, for example, regulated mid-rail termination, low power active crosstalk cancellation, circuit redundancy, per-bit duty cycle correction and skew correction (deskew), line coding, and transmitter equalization.
In one exemplary embodiment of MCPL, a PHY logic layer may be implemented, which may also facilitate (e.g., electrical layer features) extending data rates and channel configurations while also enabling the interconnect to route multiple protocols across the electrical layer. Such an implementation may provide and define a modular common physical layer that is protocol independent and designed for potentially any existing or future interconnect protocol.
Turning to fig. 6, a simplified block diagram 600 is shown representing at least a portion of a system including an exemplary embodiment of a multi-chip package link (MCPL). MCPL may use physical electrical connections to implement (e.g., wires implemented as vias) connecting a first device 605 (e.g., a first die comprising one or more sub-components) with a second device 610 (e.g., a second die comprising one or more other sub-components). In the particular example shown in the high-level representation of diagram 600, all signals (in channels 615, 620) may be unidirectional, and lanes may be provided for data signals to have both upstream and downstream data transfers. Although the block diagram 600 of fig. 6 refers to the first component 605 as an upstream component and the second component 610 as a downstream component, and the physical path of the MCPL for transmitting data as the downstream channel 615 and the path for receiving data (from the component 610) as the upstream channel 620, it should be understood that the MCPL between the devices 605, 610 may be used by each device to both transmit and receive data between the devices.
In one example implementation, the MCPL may provide a physical layer (PHY) that includes electrical MCPL PHYs 625a, 625b (or, collectively, 625) and executable logic that implements MCPL logic PHYs 630a, 630b (or, collectively, 630). The electrical or physical PHY 625 may provide a physical connection through which data is communicated to the devices 605, 610. Signal conditioning components and logic may be implemented in association with physical PHY 625 to establish high data rate and channel configuration capabilities of the link, which in some applications may involve physical connections that are closely clustered at a length of about 45mm or more. Logical PHY 630 may include logic to facilitate timing, link state management (e.g., for link layers 635a, 635 b), and potentially multiple protocol multiplexing between different protocols for communication over MCPL.
In an example embodiment, the physical PHY 625 may include, for each channel (e.g., 615, 620), a set of data paths over which in-band data may be transmitted. In this particular example, 50 data lanes are provided within upstream channel 620 and downstream channel 615, but any number of lanes may be used as permitted by layout and power constraints, desired application, device constraints, and the like. Each channel may also include one or more dedicated paths for gating, or clocks, signals for the channel, one or more dedicated paths for valid signals for the channel, one or more dedicated paths for streaming signals, and one or more dedicated paths for link state machine management or sideband signals. The physical PHY may also include sideband links 640, where in some examples, sideband links 640 may be bidirectional lower frequency control signal links to coordinate state transitions and other attributes of the MCPLs connecting devices 605, 610, among other examples.
As described above, multiple protocols may be supported using implementations of MCPL. In practice, multiple independent transaction layers 650a, 650b may be provided at each device 605, 610. For example, each device 605, 610 may support and utilize two or more protocols, e.g., PCI, PCIe, QPI, intel interconnect-in-die (IDI), and so on. The IDI is a coherency protocol on the die used to communicate between the core, Last Level Cache (LLC), memory, graphics, and IO controllers. Other protocols (including ethernet protocols, infiniband protocols, and other PCIe fabrics based on protocols) may also be supported. In other examples, the combination of logical and physical PHYs may also be used as a die-to-die interconnect to connect a SerDes PHY (PCIe, ethernet, infiniband, or other high speed SerDes) on one die to its upper layers implemented on another die.
Logical PHY 630 may support multiplexing between these multiple protocols over MCPL. For example, the dedicated flow path may be used to assert an encoded stream signal that identifies which protocol applies to data being transmitted substantially simultaneously on the data path of the channel. In addition, logical PHY 630 may be used to negotiate various types of link state transitions that various protocols may support or request. In some instances, LSM _ SB signals sent over the dedicated LSM _ SB lanes of a channel may be used with sideband link 640 to communicate and negotiate link state transitions between devices 605, 610. In addition, link training, error detection, skew detection, deskew, and other functions of the legacy interconnect may be replaced or managed, in part, using logical PHY 630. For example, the active signals transmitted in each channel through one or more dedicated active signal paths may be used for signal link activity, detecting skew, link errors, and implementing other features, among other examples. In the specific example of fig. 6, multiple active lanes are provided in each channel. For example, data lanes within a channel may be bundled or clustered (physically and/or logically) and an efficient lane may be provided for each cluster. Further, multiple gating paths may be provided, in some cases, a dedicated gating signal for each of multiple data path clusters in a channel, among other examples.
As described above, logical PHY 630 may be used to negotiate and manage link control signals sent between devices connected by MCPL. In some embodiments, logical PHY 630 may include Link Layer Packet (LLP) generation logic 660, which may be used to transmit link layer control messages over MCPL (i.e., in-band). Such messages may be sent over the data path of the channel using a flow path that identifies that data (e.g., link control data) is link-layer to link-layer messaging, among other examples. In addition to other link layer features between the link layers 635a, 635b of the devices 605, 610, link layer messages implemented using the LLP module 660 can facilitate negotiation and execution of link layer state transitions, power management, loopback, disabling, re-centering, scrambling, respectively.
Turning to fig. 7, a diagram 700 is shown representing an exemplary signaling of a set of lanes (e.g., 615, 620) in a particular channel using an exemplary MCPL. In the example of fig. 7, two clusters of twenty-five (25) data lanes are provided for fifty (50) total data lanes in the channel. A portion of the lanes are shown, while other lanes (e.g., DATA [4-46] and the second strobe signal lane (STRB)) are omitted for ease of illustration of a particular example (e.g., as redundant signals). When the physical layer is in an active state (e.g., not powered down or in a low power mode (e.g., L1 state)), a synchronous clock signal may be provided for the strobe path (STRB). In some embodiments, data may be sent on both the rising and falling edges of the strobe. Each edge (or half clock cycle) may divide a unit time interval (UI). Thus, in this example, one bit (e.g., 705) may be sent on each lane, allowing one byte to be sent every 8 UI. The byte time period 710 may be defined as 8UI, or as the time for one byte to be sent on a single one of the DATA lanes (e.g., DATA [0-49 ]).
In some implementations, VALID signals (e.g., VALID 0, VALID 1) transmitted on one or more dedicated VALID signal channels may be used as a dominant indicator for a receiving device when asserted (high) to the receiving device or receiver to identify that DATA is being transmitted on a DATA path (e.g., DATA [0-49]) from a transmitting device or source during a following time period (e.g., byte time period 710). Alternatively, when the valid signal is low, the source instructs the receiver not to transmit data on the data path during the next time period. Thus, when the receiver logic PHY detects (e.g., on lanes VALID 0 and VALID 1) that a VALID signal is not asserted, the receiver may ignore any DATA detected on the DATA lanes (e.g., DATA [0-49]) during the following time period. For example, crosstalk noise or other bits may appear on one or more of the data paths when, in fact, the source is not transmitting any data. Due to the low or non-asserted valid signal during a previous time period (e.g., a previous byte time period), the receiver may determine that the data path will be ignored during the next time period.
Data transmitted on each of the lanes of the MCPL may be tightly aligned to the gating signal. Time periods (e.g., byte time periods) may be defined based on the gating, and each of these periods may correspond to a defined window in which signals are to be sent on the DATA path (e.g., DATA [0-49]), the VALID path (e.g., VALID1, VALID2), and the flow path (e.g., STREAM). Thus, alignment of these signals may enable identification of valid signals in a previous time period window applied to data in a subsequent time period window, and identification of stream signals applied to data in the same time period window. The stream signal may be an encoded signal (e.g., 1 byte of data for a byte time period window) encoded to identify a protocol that is applied to data transmitted during the same time period window.
To illustrate, in the particular example of FIG. 7, a byte time period window is defined. Before any DATA is injected on DATA path DATA [0-49], a valid is asserted for a window of time periods n (715). At a subsequent window of time periods n +1(720), data is transmitted on at least some of the data lanes. In this case, data is sent on all fifty data lanes during n +1 (720). Because the assertion was valid for the duration of the previous time period window n (715), the receiver device may verify the DATA received on DATA lanes DATA [0-49] during time period windows n +1 (720). Additionally, the dominant nature of the valid signal during time period window n (715) allows the receiving device to prepare the incoming data. Continuing with the example of FIG. 7, the VALID signal remains asserted (on VALID1 and VALID2) during time period window n +1(720), so that the receiver device expects to send DATA over DATA paths DATA [0-49] during time period window n +2 (725). The receiver device may also expect to receive (and process) additional data transmitted during the immediately following time period window n +3(730) if the valid signal is to remain asserted during time period window n +2 (725). However, in the example of FIG. 7, the valid signal is deasserted for the duration of time period window n +2(725), indicating that no DATA will be sent by the receiver device during time period window n +3(730), and that any bits detected on DATA lanes DATA [0-49] during time period window n +3(730) should be ignored.
As described above, a plurality of active paths and gated paths may be maintained at each channel. This may help, among other advantages, maintain circuit simplicity and synchronization among clusters of relatively long physical paths connecting two devices. In some embodiments, the set of data lanes may be divided into clusters of data lanes. For example, in the example of FIG. 7, the DATA lanes DATA [0-49] may be divided into two clusters of twenty-five lanes and each cluster may have a dedicated active and strobe lane. For example, VALID path VALID1 may be associated with DATA paths DATA [0-24] and VALID path VALID2 may be associated with DATA paths DATA [25-49 ]. The signal on each "duplicate" path of the active and gated paths for each cluster may be the same. The VALID and STREAM lanes may also be different between clusters, allowing them to be operated as independent ports in the fabric or as reduced width ports to save power.
As described above, DATA on the flow path STREAM may be used to indicate the receive logic PHY of the protocol applied to the corresponding DATA being transmitted on the DATA path DATA [0-49 ]. In the example of FIG. 7, the STREAM signal is sent on the STREAM during the same window of time period as the DATA on DATA lanes DATA [0-49] to indicate the protocol of the DATA on the DATA lanes. In an alternative embodiment, the stream signal (e.g., along with the corresponding valid signal, among other potential modifications) may be transmitted during a previous time period window. However, continuing with the example of FIG. 7, during time period window n +2(725), stream signal 735 is sent, which stream signal 735 is encoded to indicate the protocol (e.g., PCIe, PCI, IDI QPI, etc.) that is to be applied to the bits sent over DATA lanes DATA [0-49] during time period window n +2 (725). Similarly, another stream signal 740 may be transmitted during a subsequent time period window n +2(725), to indicate the protocol applied to the bits transmitted over DATA lanes DATA [0-49] during time period window n +2(725), and so on. In some cases, for example, the example of fig. 7 (where both stream signals 735, 740 have the same encoding, binary FF), the data in sequential time period windows (e.g., n +1(720) and n +2(725)) may belong to the same protocol. However, in other cases, the DATA in sequential time period windows (e.g., n +1(720) and n +2(725)) may come from different transactions applied by different protocols, and the stream signals (e.g., 735, 740) may be encoded to identify different protocols applied to sequential bits of DATA on the DATA lanes (e.g., DATA [0-49]), among other examples.
In some embodiments, a low power state or idle state may be defined for MCPL. For example, the physical layer (electrical and logical) of the MCPL may enter an idle state or a low power state when no device is transmitting data on the MCPL. For example, in the example of fig. 7, at time period window n-2(745), MCPL is in a quiet or idle state and gating is disabled to save power. The MCPL may transition out of the low power mode or idle mode, waking up the strobe at time period window n-1 (e.g., 705). The gating may complete transmission of the preamble (e.g., to facilitate each of the lanes of the wakeup and synchronization channel and the receiver device), starting the gated signal before any other signals on the other ungated lanes are transmitted. After the time period window n-1(705), a valid signal may be asserted for time period window n (715) to inform the receiver that data is present in the next time period window n +1(720) discussed above.
The MCPL may re-enter a low power state or idle state (e.g., L1 state) upon detection of other lane idle conditions with respect to active lanes, data lanes, and/or MCPL channels. For example, no signaling may be detected starting at time period window n +3(730) and onwards. Logic on the source or sink device may initiate a transition back to a low power state, again resulting (e.g., time period window n +5(755)) in entering an idle strobe in a power save mode, among other examples and principles, including those discussed later herein.
The electrical characteristics of the physical PHY may include one or more of the following: single-ended signaling, half-rate forward timing, matching between the on-chip propagation delays of the interconnect channels and transmitters (sources) and receivers (receivers), optimized electrostatic discharge (ESD) protection, pad capacitance, and other features. In addition, MCPL may be implemented to achieve higher data rate (e.g., near 16 Gb/sec) and energy efficiency characteristics than conventional package I/O solutions.
Fig. 8 shows a portion of a simplified block diagram 800 representing a portion of an exemplary MCPL. The diagram 800 of fig. 8 includes a representation of an exemplary path 805 (e.g., a data path, an active path, or a flow path) and clock generation logic 810. As shown in the example of fig. 8, in some embodiments, the clock generation logic 810 may be implemented as a clock tree for distributing the generated clock signal to each block of each lane (e.g., data lane 805) implementing the exemplary MCPL. In addition, a clock recovery circuit 815 may be provided. In some embodiments, rather than providing a separate clock recovery circuit for each lane in which the clock signal is distributed, as is conventional in at least some conventional interconnect I/O architectures, a separate clock recovery circuit may be provided to a cluster of multiple lanes. Indeed, when applied to the exemplary configurations in fig. 6 and 7, a separate strobe path and accompanying clock recovery circuitry may be provided for each cluster of twenty-five data paths.
Continuing to the example of fig. 8, in some embodiments, at least the data path, the flow path, and the active path may terminate at the intermediate rail at a regulated voltage greater than zero (ground). In some embodiments, the mid-rail voltage may be regulated to VCC/2. In some implementations, separate voltage regulators 825 may be provided at each cluster of lanes. For example, when applied to the examples of fig. 6 and 7, among other potential examples, the first voltage regulator may be provided to a first cluster of twenty-five data lanes and the second voltage regulator may be provided to the remaining cluster of twenty-five data lanes. In some cases, the example voltage regulator 825 may be implemented as a linear regulator, switched capacitor circuit, among other examples. In some embodiments, an analog feedback loop or a digital feedback loop may be provided for the linear regulator, among other examples.
In some embodiments, crosstalk cancellation circuitry may also be provided to the exemplary MCPL. In some cases, the tight nature of long MCPL wires can introduce crosstalk interference between vias. Crosstalk cancellation logic may be implemented to address these and other issues. For example, in one example shown in fig. 9-10, crosstalk can be significantly reduced with low power active circuits as shown in fig. 900 and 1000. For example, in the example of fig. 9, a weighted high-pass filtered "aggressor" signal may be added to a "victim" signal (i.e., the signal suffers crosstalk interference from aggressors). Each signal may be considered a victim of crosstalk of each other signal in the link and may itself be an aggressor to another signal, as long as it is a source of crosstalk interference. Due to the derivative nature of crosstalk on the link, such signals can be generated and reduced by over 50% of the crosstalk on the victim path. In the example of fig. 9, the low-pass filtered aggressor signal may be generated by a high-pass RC filter (e.g., implemented by C and R1) that produces a filtered signal to be added using a summing circuit 905 (e.g., an RX sense-amplifier).
While embodiments of the circuit may be implemented with relatively low overhead, as illustrated in the diagram of fig. 10, which illustrates an exemplary transistor-level circuit diagram of the circuit shown and described in the example of fig. 9, embodiments similar to those described in the example of fig. 9 may be a particularly convenient solution for applications (e.g., MCLP). It should be understood that the representations in fig. 9 and 10 are simplified representations, and that practical implementations will include multiple copies of the circuits shown in fig. 9 and 10 to accommodate networks of crosstalk interference among and between the paths of the links. As an example, in a three-way link (e.g., ways 0-2), a circuit similar to that described in the examples of fig. 9 and 10, in other examples, may be provided from way 0 to way 1, from way 0 to way 2, from way 1 to way 0, from way 1 to way 2, from way 2 to way 0, from way 2 to way 1, and so on, based on the geometric layout of the ways.
Additional features may be implemented at the physical PHY level of the exemplary MCPL. For example, in some cases, receiver offsets may introduce significant errors and limit I/O voltage margins. Circuit redundancy may be used to improve the sensitivity of the receiver. In some embodiments, circuit redundancy may be optimized to account for standard deviation offset of the data sampler used at the MCPL. For example, an exemplary data sampler that may be provided is designed to a three (3) standard deviation offset specification. In the examples of fig. 6 and 7, for example, two (2) data samplers would be used for each receiver (e.g., for each lane), and one hundred (100) samplers would be used for fifty (50) lanes of MCPL. In this example, the probability of one of the Receiver (RX) paths failing the three standard deviation offset specification is 24%. If another one of the other data samplers is found to exceed the limit, a chip reference voltage generator may be provided to set an upper offset limit and move to the next data sampler on the receiver. However, each receiver uses four (4) data samplers (i.e., instead of two in this example), and the receiver will fail if three-quarters of the samplers fail. For fifty lanes MCPL, as in the examples of fig. 6 and 7, adding this additional circuit redundancy can significantly reduce the failure rate from 24% to less than 0.01%.
In other examples, at very high data rates, per-bit Duty Cycle Correction (DCC) and skew correction may be used to increase the baseline for each cluster of DCC and skew correction to improve link margin. Instead of correcting for all cases (as in conventional solutions), in some embodiments, a low power digital implementation may be used to sense and correct outliers in which the I/O lines fail. For example, a global adjustment of the path may be performed to identify a problem path within the cluster. These problem paths can then be targeted for each path adjustment to achieve the high data rates supported by the MCPL.
Additional features may also optionally be implemented in some instances of MCPL to improve the performance characteristics of the physical link. For example, line coding may be provided. Although mid-rail termination as described above may allow the DC Data Bus Inversion (DBI) to be omitted, AC DBI may still be used to reduce dynamic power. More complex encoding may be used to eliminate worst case differences of 1 and 0 to reduce driver requirements, e.g., mid-rail regulators, and to limit I/O switching noise, among other example benefits. Transmitter equalization may also optionally be implemented. For example, at very high data rates, insertion loss can be significant for channels within the package. In some cases, two-tap weight transmitter equalization (e.g., performed during an initial power-up sequence) may be sufficient to mitigate some of these problems, and so on.
Turning to fig. 11, a simplified block diagram 1001 illustrates an exemplary logical PHY showing an exemplary MCPL. Physical PHY 1105 may be connected to a die that includes logical PHY1110 and additional logic to support the link layer of MCPL. In this example, the die may also include logic to support a plurality of different protocols over the MCPL. For example, in the example of fig. 11, PCIe logic 1115 and IDI logic 1120 may be provided such that the dies may communicate using PCIe or IDI at the same MCPL connecting the two dies, including in potentially many other examples where more than two protocols are supported by the MCPL or protocols other than PCIe and IDI are supported by the MCPL. Various protocols supported between dies may provide varying levels of services and features.
The logic PHY1110 may include logic to negotiate link state transitions (e.g., received over PCIe or IDI) related to requests by upper layer logic of the die. In some implementations, the logic PHY1110 may also include link test and debug logic (e.g., 1130). As described above, an exemplary MCPL may support control signals that are sent between dies through the MCPL to facilitate protocol agnostic, high performance, and power efficiency characteristics (among other example characteristics) of the MCPL. For example, as described in the above examples, logical PHY1110 may support the generation, transmission, and reception of valid, streaming, and LSM sideband signals in connection with transmitting and receiving data over a dedicated data path.
In some implementations, multiplexing logic (e.g., 1135) and demultiplexing logic (e.g., 1140) may be included in logical PHY1110 or otherwise accessible to logical PHY 1110. For example, multiplexing logic (e.g., 1135) can be used to identify data (e.g., packets, messages, etc. in particular) that is to be sent out to the MCPL. Multiplexing logic 1135 may identify the protocol of the management data and generate a stream signal encoded to identify the protocol. For example, in one exemplary embodiment, the stream signal may be encoded as two hexadecimal symbol bytes (e.g., IDI: FFh; PCIe: F0 h; LLP: AAh; sideband: 55h, etc.) and may be transmitted during the same window of data (e.g., byte time cycle window) managed by the identified protocol. Similarly, demultiplexing logic 1140 may be used to interpret incoming stream signals to decode the stream signals and identify protocols for application to data received concurrently with the stream signals on the data path. De-multiplexing logic 1140 may then apply (or ensure) the link layer processing of the particular protocol and cause the data to be processed by the corresponding protocol logic (e.g., PCIe logic 1115 or IDI logic 1120).
The logic PHY1110 may also include link layer grouping logic 1150 that may be used to handle various link control functions including power management tasks, loopback, disabling, re-centering, scrambling, and the like. Among other functions, the LLP logic 1150 may facilitate link layer to link layer messaging through MCLP. Data corresponding to LLP signaling may also be identified by a stream signal transmitted on a dedicated stream signal path that is encoded to identify data path LLP data. Multiplexing logic and demultiplexing logic (e.g., 1135, 1140) may also be used to generate and interpret stream signals corresponding to LLP traffic, and to cause such traffic to be processed by suitable die logic (e.g., LLP logic 1150). Likewise, in other examples, some embodiments as MCLPs may include dedicated sidebands (e.g., sidebands 1155 and support logic), such as asynchronous and/or lower frequency sideband channels.
The logical PHY logic 1110 may also include state machine management logic that may generate and receive (and use) link state management messages through the dedicated LSM sideband pathway. For example, in other potential examples, the LSM sideband pathway may be used to perform a feedback check (handsign) to advance the link training state, exiting from the power management state (e.g., the L1 state). In other examples, the LSM sideband signal may be an asynchronous signal because it is not aligned with the data, valid stream signals of the link, but rather corresponds to signaling state transitions and aligning the link state machine between two dies or chips connected by the link. In some examples, providing a dedicated LSM sideband pathway may allow detection circuitry of traditional squelch and Analog Front End (AFE) reception to be eliminated, among other example benefits.
Turning to fig. 12, a simplified block diagram 1200 illustrates another representation showing logic for implementing MCPL. For example, a defined Logical PHY Interface (LPIF)1205 is provided for the logical PHY1110 by any of a number of different protocols (e.g., PCIe, IDI, QPI, etc.) 1210, 1215, 1220, 1225, and signaling modes (e.g., sidebands) may interface with the physical layer of the exemplary MCPL. In some embodiments, multiplexing and arbitration logic 1230 may also be provided as a separate layer from logical PHY 1110. In one example, the LPIF 1205 may be provided as an interface on either side of the MuxArb layer 1230. Logical PHY1110 may interface with a physical PHY (e.g., Analog Front End (AFE)1105 of an MCPL PHY) via another interface.
The LPIF may extract PHYs (logical and electrical/analog) from upper layers (e.g., 1210, 1215, 1220, 1225) so that a completely different PHY may be implemented under the LPIF transparent to the upper layers. In other examples, this may help to promote modularity and re-use in design, such as when the underlying signaling technology PHY is updated, the upper layers may remain intact. In addition, the LPIF may define a plurality of signals implementing multiplexing/demultiplexing, LSM management, error detection and processing, and other functions of the logical PHY. For example, table 1 summarizes at least a portion of the signals that may be defined for an exemplary LPIF:
table 1:
TABLE 1
As shown in Table 1, in some embodiments, an alignment mechanism may be provided by the AlignReq/AlignAck feedback test. For example, some protocols may lose packet framing when the physical layer enters recovery. For example, to ensure proper framing identification by the link layer, alignment of packets may be corrected. Further, as shown in fig. 13, when the physical layer enters recovery, the physical layer may assert the StallReq signal so that the link layer asserts a stall signal when a new aligned packet is ready to be transmitted. The physical layer logic may sample both stall (stall) and Valid (Valid) to determine whether the packet is aligned. For example, the physical layer may continue to drive trdy to drain link layer packets until sampled stall (stall) and valid are asserted, including using valid other alternative implementations for assisting in packet alignment in other potential implementations.
Various fault tolerances may be defined for signaling on the MCPL. For example, fault tolerance may be defined for active, streaming, LSM sideband, low frequency sideband, link layer packet, and other types of signals. The fault tolerance for packets, messages, and other data sent over the MCPL's dedicated data path may be based on the particular protocol that manages the data. In some embodiments, error detection and handling mechanisms may be set, such as Cyclic Redundancy Check (CRC), retry buffer, among other potential examples. As an example, for PCIe packets sent over MCPL, a 32-bit CRC may be used for PCIe Transaction Layer Packets (TLPs) with guaranteed delivery (e.g., over a replay mechanism) and a 16-bit CRC may be used for PCIe link layer packets (which may be constructed to be lossy (e.g., where replay is applied)). Further, for PCIe framing tokens, a particular hamming distance (e.g., a hamming distance of four (4)) may be defined for the token identifier, among other examples; parity and 4-bit CRC may also be utilized. In another aspect, for an IDI packet, a 16-bit CRC may be utilized.
In some embodiments, fault tolerance may be defined for Link Layer Packets (LLPs) that include a transition (i.e., 0 to 1) that requires a valid signal from low to high (e.g., to help ensure bit and symbol lock). Further, in one example, in other defined characteristics that can be used as a basis for determining a failure in LLP data on MCPL, with a requester retry after a response timeout, a certain number of consecutive, identical LLPs can be defined to be sent, and a response can be expected for each request. In a further example, fault tolerance may be provided to the valid signal, for example, by extending the valid signal, or symbol, over the entire time period window (e.g., by keeping the valid signal high for eight UIs). Further, errors or faults in the stream signal may be prevented by maintaining a hamming distance for the encoded values of the stream signal, among other examples.
Embodiments of the logical PHY may include error detection, error reporting, and error handling logic. In some implementations, the logical PHY of the example MCPL may include, among other examples, logic to detect PHY layer de-framing errors (e.g., on the active and flow path), sideband errors (e.g., with respect to LSM state transitions), errors in the LLP (e.g., critical to LSM state transitions). Among other examples, some error detection/resolution may be passed to upper layer logic, such as PCIe logic adapted to detect specific PCIe errors.
In the case of de-framing errors, in some embodiments, one or more mechanisms may be provided by the error handling logic. De-framing errors may be handled based on the protocol involved. For example, in some embodiments, the link layer may be notified of an error to trigger a retry. De-framing may also cause the logical PHY to de-align the frames. Further, among other techniques, re-centering of the logical PHY may be performed and the symbol/window lock may be reacquired. In some examples, centering may include the PHY shifting the receiver clock phase to an optimal point to detect incoming data. In this context, "best" may refer to where it has the greatest tolerance for noise and clock jitter. Among other examples, the re-centering may include a simplified centering function that is performed, for example, when the PHY wakes up from a low power state.
Other types of errors may involve other error handling techniques. For example, errors detected at the sidebands may be captured by a timeout mechanism (e.g., of the LSM) of the corresponding state. The error may be logged and then the link state machine may be transitioned to reset. The LSM may remain in the reset state until a restart command is received from the software. In another example, an LLP error (e.g., a link control packet error) can be handled using a timeout mechanism that can restart the LLP sequence if an acknowledgement to the LLP sequence is not received.
Fig. 14A-14C show representations of exemplary bit maps on the data path of an exemplary MCPL for various types of data. For example, an exemplary MCPL may include fifty data lanes. Fig. 14A shows a first bit map (such as an IDI) of an exemplary 16-byte slot in a first protocol, which may be sent over a data lane within an 8UI symbol or window. For example, within a defined 8UI window, three 16 byte slots (including header slots) may be sent. In this example, two bytes of DATA are held, and these remaining two bytes may be utilized as CRC bits (e.g., DATA [48] and DATA [49] on the lane).
In another example, fig. 14B illustrates a second exemplary bitmap for PCIe packet data sent over the fifty data lanes of the exemplary MCPL. In the example of fig. 14B, a 16 byte packet (e.g., a Transaction Layer (TLP) or a Data Link Layer (DLLP) PCIe packet) may be sent through the MCPL. In an 8UI window, three packets may be sent with the remaining two bytes within the window leaving unused bandwidth. Framing tokens may be included within the symbols and used to locate the start and end of each packet. In one example of PCIe, the framing used in the example of fig. 14B may be the same as those tokens implemented at 8 GT/sec PCIe.
In yet another example, as shown in fig. 14C, an exemplary bitmap of link-to-link packets (e.g., packet LLPs) transmitted over an exemplary MCPL is shown. The LLPs may be every 4 bytes, and each LLP (e.g., LLP0, LLP1, LLP2, etc.) may be transmitted four times in succession according to fault tolerance and error detection within the exemplary embodiment. For example, failure to receive four consecutive identical LLPs may indicate an error. Additionally, failure to receive a VALID within a travel time window or symbol may also indicate an error when of other data types. In some cases, the LLP may have a fixed slot. Further, in this example, among other examples, unused, or "spare," bits in a byte time period may result in a logic 0 being transmitted over two of the fifty lanes (e.g., DATA [48-49 ]).
Turning to fig. 15, a simplified link state machine transition diagram 1400 is shown along with sideband feedback check exchanges utilized between state transitions. For example, the reset.idle state (e.g., where Phase Locked Loop (PLL) lock calibration is performed) may be converted to the reset.cal state (e.g., where the link is further calibrated) by sideband feedback verification. Cal can be converted to a reset.clockdcc state by sideband feedback verification (e.g., where Duty Cycle Correction (DCC) and Delay Locked Loop (DLL) locking can be performed). Additional feedback checks may be performed to transition from reset.clockdcc to reset.quiet state (e.g., to de-assert the valid signal). To assist in alignment of signaling on the path of the MCPL, the path may be centered by the center.
In some embodiments, as shown in the example of fig. 16, during the center. The receiver may adjust its receiver circuitry to receive this training pattern, for example, by setting the phase interpolator position and vref position and setting the comparator. The receiver may continually compare the received pattern to the expected pattern and store the result in a register. After a set of modes is completed, the receiver may increase the phase interpolator setting to keep vref the same. The test pattern generation and comparison process may continue and new comparison results may be stored in the registers using a process that repeats steps through all phase interpolator values and through all values of vref. The query state may be entered when the pattern generation and comparison processes are all completed. The path is hereinafter centered by the center.pattern and Center query link states, and sideband feedback checks (which may, for example, use LSM band signals of a dedicated LSM sideband path through the link) facilitate transitions to the link.init state to initialize the MCPL and enable transmission of data on the data MCPL.
Returning briefly to the discussion of FIG. 15, as described above, sideband feedback verification may be used to facilitate link state machine transitions between dies or chips in a multi-chip package. For example, the signal on the LSM sideband path of the MCPL may be used to synchronize state machine transitions throughout the die. For example, when the conditions for the exit state (e.g., reset. idle) are satisfied, the edge that satisfies these conditions may assert the LSM sideband signal on its outbound LSM _ SB path and wait for other remote dies to reach the same conditions and assert the LSM sideband signal on its LSM _ SB path. When both LSM _ SB signals are asserted, the link state machine of each respective die may transition to the next state (e.g., reset. A minimum overlap time may be defined during which both LSM _ SB signals should remain asserted before transitioning states. Further, after LSM _ SB is deasserted, a minimum dwell time may be defined to allow accurate turn detection. In some embodiments, each link state machine transition may be conditioned on and facilitated by such LSM _ SB feedback checks.
Fig. 17 is a more detailed link state machine diagram 1700 illustrating at least some of the additional link states and link state transitions that may be included in the exemplary MCPL. In some embodiments, among other states and state transitions as shown in fig. 17, an exemplary link state machine may include a "directed loop back" transition may be provided to place the path of the MCPL in a digital loop back. For example, after the clock recovery circuit, the receiver path of the MCPL may be looped back to the transmitter path. The "LB register" state may also be provided that may be used to align the digital symbols in some cases. Further, as shown in fig. 15, MCPL may support, among potentially other examples, multiple link states, including an active L0 state and a low power state (e.g., an L1 idle state, and an L2 sleep state).
FIG. 18 is a simplified block diagram 1800 illustrating an exemplary flow of transitioning between an active state (e.g., L0) and a low power, or idle, state (e.g., L1). In this particular example, a first device 1805 and a second device 1810 are communicatively coupled using MCPL. When active, DATA is transferred through the MCPL's lanes (e.g., DATA, VALID, STREAM, etc.). Link Layer Packets (LLPs) may pass through lanes (e.g., data lanes with flow signals indicating that the data is LLP data) to assist in facilitating link state transitions. For example, LLPs may be sent between the first device 1805 and the second device 1810 to negotiate from L0 into L1. For example, an upper layer protocol supported by the MCPL may communicate a need to enter L1 (or another state) and the upper layer protocol may cause an LLP to be sent over the MCPL to facilitate link layer feedback checking to cause the physical layer to enter L1. For example, fig. 18 shows that sending at least a portion of the LLPs includes an "enter L1" request LLP sent from the second (upstream) device 1810 to the first (downstream) device 1805. In some embodiments, and in upper layer protocols, the downstream port does not initiate entry into L1. Among other examples, the receiving first device 1805 may send a "change to L1" request LLP in response, where the second device 1810 may acknowledge an Acknowledgement (ACK) LLP by "change to L1". Upon detecting that the feedback check is complete, the logical PHY may cause the sideband link to be asserted on the dedicated sideband link to acknowledge that the ACK was received and that the device (e.g., 1805) is ready and expecting to enter L1. For example, the first device 1805 may assert a sideband signal 1815 sent to the second device 1810 to acknowledge receipt of the last ACK at the link layer feedback check. Additionally, the second device 1810 may also assert a sideband signal in response to the sideband signal 1815 to notify the first device 1805 of the sideband ACK 1805 of the first device. With link layer control and sideband feedback verification completed, the MCPL PHY can be transitioned to the L1 state such that all lanes of the MCPL are placed in idle power saving mode, including the MCPL gating 1820, 1825 of the respective device 1805, 1810. L1 may be exited after upper level logic of one of the first device 1805 and the second device 1810 requests re-entry into L0 (e.g., in response to detecting that data is to be sent to the other device through the MCPL).
As described above, in some embodiments, the MCPL may facilitate communication between two devices that support potentially multiple different protocols, and the MCPL may facilitate communication according to potentially any of the multiple protocols on the path of the MCPL. However, facilitating multiple protocols may duplicate entry and re-entry into at least some of the linked states. For example, while some conventional interconnects have a single upper layer protocol that assumes the role of a host in state transitions, implementations of MCPL with multiple different protocols effectively involve multiple hosts. As an example, as shown in fig. 18, each of PCIe and IDI may be supported between two devices 1805, 1810 through an implementation of MCPL. For example, placing the physical layer into an idle or low power state may be regulated according to first being granted permission from each of the supported protocols (e.g., PCIe and IDI).
In some cases, entry into L1 (or another state) may be requested by supporting only one of a plurality of supported protocols for implementation of MCPL. While there may be a possibility that other protocols will likewise request entry into the same state (e.g., based on identifying similar conditions on the MCPL (e.g., little or no traffic)), the logical PHY may wait until a grant or instruction is received from each upper layer protocol before actually facilitating the state transition. The logical PHY may track which upper layer protocols have requested a state change (e.g., perform a corresponding feedback check) and trigger a state transition (e.g., a transition from L0 to L1 or another transition that would affect or interfere with other protocol communications) when each of the identified protocols has requested a particular state change. In some embodiments, the protocols may be blinded (blind) for their dependence at least in part on other protocols in the system. Further, in some cases, a protocol may expect a response to a request (e.g., from a PHY) to enter a particular state, e.g., an acknowledgement or rejection of a requested state transition. Thus, in this case, while waiting for permission from other supported protocols to enter the idle link state, the logical PHY may generate a synthetic response to the request to enter the idle state to "fool" the requesting upper layer protocol into believing that a particular state has been entered (in fact, when the path is still active, at least until other protocols also request entry into the idle state). Among other potential advantages, this may simplify coordinating entry into low power states between multiple protocols in other examples.
It should be noted that the above described apparatus, methods and systems may be implemented in any electronic device or system as described above. By way of specific illustration, the following figures provide exemplary systems for utilizing the invention described herein. As the following system is described in greater detail, a number of interconnects that differ from the discussion above are disclosed, described, and re-discussed. And it will be apparent that the above-described advances may be applied to one or more of these interconnects, structures or architectures.
Referring to FIG. 19, an embodiment of a block diagram for a computing system including a multicore processor is depicted. Processor 1900 includes any processor or processing device for executing code, such as a microprocessor, embedded processor, Digital Signal Processor (DSP), network processor, hand-held processor, application processor, co-processor, system on a chip (SOC), or other device. In one embodiment, processor 1900 includes at least two cores, core 1901 and core 1902, which may include asymmetric cores or symmetric cores (as shown in the embodiments). However, processor 1900 may include any number of processing elements that may be symmetric or asymmetric.
In one embodiment, a processing element refers to hardware or logic to support a software thread. Examples of hardware processing elements include: a thread unit, a thread slot, a thread, a processing unit, a context unit, a logical processor, a hardware thread, a core, and/or any other element capable of maintaining a state (e.g., an execution state or an architectural state) for a processor. In other words, in one embodiment, a processing element refers to any hardware capable of being independently associated with code (e.g., software threads, operating systems, applications, or other code). A physical processor (or processor socket) generally refers to an integrated circuit, which may include any number of other processing elements (e.g., cores or hardware threads).
A core typically refers to logic located on an integrated circuit capable of maintaining an independent architectural state, where each independently maintained architectural state is associated with at least some dedicated execution resources. In contrast to cores, a hardware thread generally refers to any logic located on an integrated circuit capable of maintaining an independent architectural state, wherein the independently maintained architectural states share access to execution resources. It can be seen that the naming of the hardware threads and the lines between cores overlap when some resources are shared and others are dedicated to the architectural state. Typically, however, the cores and hardware threads are viewed by the operating system as independent logical processors, where the operating system is able to independently schedule operations on each logical processor.
As shown in FIG. 19, physical processor 1900 includes two cores — core 1901 and core 1902. Here, core 1901 and core 1902 are considered symmetric cores, i.e., cores having the same configuration, functional units, and/or logic. In another embodiment, core 1901 comprises an out-of-order processor core, while core 1902 comprises an in-order processor core. However, core 1901 and core 1902 may be individually selected from any type of core (e.g., a native core, a software managed core, a core adapted to execute a native Instruction Set Architecture (ISA), a core adapted to execute a translated Instruction Set Architecture (ISA), a co-designed core, or other known core). In a heterogeneous core environment (i.e., asymmetric cores), some form of translation (e.g., binary translation) may be utilized to schedule or execute code on one or both cores. For further discussion, the functional units shown in core 1901 are described in more detail below, while the units in core 1902 operate in a similar manner in the described embodiments.
As shown, core 1901 includes two hardware threads 1901a and 1901b, which may also be referred to as hardware thread slots 1901a and 1901 b. Thus, in one embodiment, a software entity such as an operating system potentially views processor 1900 as four separate processors, i.e., four logical processors or processing elements capable of executing four software threads concurrently. As described above, a first thread is associated with architecture state registers 1901a, a second thread is associated with architecture state registers 1901b, a third thread may be associated with architecture state registers 1902a, and a fourth thread may be associated with architecture state registers 1902 b. Here, as described above, each of the architectural status registers (1901a, 1901b, 1902a, and 1902b) may be referred to as a processing element, a thread slot, or a thread unit. As shown, the architecture state registers 1901a are replicated in the architecture state registers 1901b, so independent architecture state/context can be stored for the logical processor 1901a and the logical processor 1901 b. In core 1901, other smaller resources (e.g., instruction pointers and rename logic in allocator and renamer block 1930) may also be replicated for threads 1901a and 1901 b. Some resources (e.g., reorder buffers in reorder/retirement unit 1935, ILTB1920, load/store buffers, and queues) may be shared through the partitions. Other resources (e.g., general purpose internal registers, page table base register(s), lower level data cache and data-TLB 1915, execution unit(s) 1940, and portions of out-of-order unit 1935) may be shared entirely.
The core 1901 also includes a decode module 1925 coupled to the fetch unit 1920 to decode the fetched elements. In one embodiment, the fetch logic includes separate sequencers associated with the thread slots 1901a, 1901b, respectively. Generally, core 1901 is associated with a first ISA that defines/specifies instructions executable on processor 1900. Typically, machine code instructions that are part of the first ISA typically include a portion of the instruction (referred to as an opcode), which references/specifies an instruction or operation to be performed. Decode logic 1925 includes circuitry to recognize instructions from their opcodes and pass the decoded instructions through the pipeline for processing as defined by the first ISA. For example, as discussed in more detail below, in one embodiment, the decoder 1925 includes logic designed or adapted to identify a particular instruction, such as a transactional instruction. As the decoder 1925 recognizes, the architecture or core 1901 takes specific, predefined actions to perform the task associated with the appropriate instruction. It is important to note that one or more of the tasks, blocks, operations, and methods described herein may be performed in response to a single or multiple instructions; some of these instructions may be new or old instructions. In one embodiment, it should be noted that decoder 1926 recognizes the same ISA (or a subset thereof). Alternatively, in a heterogeneous core environment, decoder 1926 recognizes a second ISA (either the first ISA or a subset of a different ISA).
In one example, allocator and renamer block 1930 includes an allocator to reserve resources (e.g., a register file to store instruction processing results). However, threads 1901a and 1901b may be capable of out-of-order execution, with allocator and renamer block 1930 also reserving other resources, such as reorder buffers for tracking instruction results. Unit 1930 may also include a register renamer to rename program/instruction reference registers to other registers internal to processor 1900. Reorder/retirement unit 1935 includes components to support out-of-order execution and later in-order retirement of instructions executed out-of-order, such as the reorder buffers, load buffers, and store buffers mentioned above.
In one embodiment, scheduler(s) and execution unit block 1940 includes a scheduler unit to schedule instructions/operations on execution units. For example, a floating point instruction is scheduled on a port of an execution unit that has an available floating point execution unit. A register file associated with the execution unit is also included to store information instruction processing results. Exemplary execution units include floating point execution units, integer execution units, jump execution units, load execution units, store execution units, and other known execution units.
The lower level data cache and data translation buffer (D-TLB)1950 is coupled to execution unit(s) 1940. The data cache is used to store recently used/operated on elements, such as data operands, which may be held in a memory coherency state. The D-TLB is to store recent virtual/linear to physical address translations. As a specific example, a processor may include a page table structure to break up physical memory into a plurality of virtual pages.
Here, core 1901 and core 1902 share a higher level or further away (fuse-out) cache, e.g., a second level cache associated with on-chip interface 1910. It should be noted that higher or further away refers to the cache level increasing or further away from the execution unit(s). In one embodiment, the higher level cache is the last level data cache, the last cache in the memory hierarchy on processor 1900. However, higher levels of cache are not so limited, and may be associated with or include an instruction cache. A trace cache, an instruction cache, may instead be coupled after decoder 1925 to store recently decoded traces. Here, an instruction may refer to a macro-instruction (i.e., a general-purpose instruction recognized by a decoder) that may be decoded into a plurality of micro-instructions (micro-operations).
In the depicted configuration, the processor 1900 also includes an on-chip interface module 1910. Historically, memory controllers, described in more detail below, have been included in computing systems external to processor 1900. In this case, on-chip interface 1910 is used to communicate with devices external to processor 1900, such as a system memory 1975, a chipset (typically including a memory controller hub for connecting to memory 1975 and an I/O controller hub for connecting to peripherals), a memory controller hub, a Northbridge, or other integrated circuits. And in this case, bus 1905 may include any known interconnect, such as a multi-drop bus, a point-to-point interconnect, a serial interconnect, a parallel bus, a coherent (e.g., cache coherent) bus, a layered protocol architecture, a differential bus, and a GTL bus.
The memory 1975 may be dedicated to the processor 1900 or shared with other devices in the system. Examples of common types of memory 1975 include DRAM, SRAM, non-volatile memory (NV memory), and other known storage devices. It should be noted that device 1980 may include a graphics accelerator, processor or card coupled to a memory controller hub, data storage coupled to an I/O controller hub, a wireless transceiver, a flash memory device, an audio controller, a network controller or other known device.
Recently, however, as more logic and devices are integrated on a separate die, such as an SOC, each of these devices may be incorporated into the processor 1900. For example, in one embodiment, the memory controller is centralized on the same package and/or on a die with the processor 1900. Here, a portion 1910 of the core (on-core) includes one or more controllers for interfacing with other devices (e.g., memory 1975 or graphics device 1980). The configuration that includes the interconnect and the controller for interfacing with such devices is commonly referred to as an on-core (or un-core) configuration. By way of example, on-chip interface 1910 includes a ring interconnect for on-chip communications and a high-speed serial point-to-point link 1905 for off-chip communications. However, in an SOC environment, even more devices such as network interfaces, coprocessors, memory 1975, graphics processor 1980, and any other known computer devices/interfaces may be integrated on separate dies or integrated circuits to provide a small form factor with high functionality and low power consumption.
In one embodiment, the processor 1900 is capable of executing compiler, optimization, and/or translator code 1977 to compile, translate, and/or optimize application code 1976 to support or interface with the apparatus and methods described herein. A compiler typically comprises a program or set of programs that convert source text/source code into target text/target code. Typically, compiler/application code is done in multiple stages and passes using a compiler to convert high-level programming language code to low-level machine or assembly language code. However, a one-way compiler can still be used for simple compilation. The compiler may utilize any known compilation technique and perform any known compiler operation, such as lexical analysis, preprocessing, parsing, semantic analysis, code generation, code conversion, and code optimization.
Larger compilers typically include multiple phases, but most typically these phases are included in two general phases: (1) a front-end, i.e., where syntactic processing, semantic processing, and some transformations/optimizations may typically occur, and (2) a back-end, i.e., where analysis, transformation, optimization, and code generation typically occur. Some compilers refer to a middle segment, which illustrates the blurring of delineation between the front-end and the back-end of the compiler. Thus, references to insertion, association, generation, or other operations of the compiler may occur at any of the stages or in any of the processes (pass) described above and any other known stages or processes of the compiler. As an illustrative example, a compiler might insert an operation, call, function, etc. in one or more stages of compilation, e.g., insert a call/operation in a front-end stage of compilation, and then convert the call/operation to lower-level code during a conversion stage. It should be noted that during dynamic compilation, compiler code or dynamic optimization code may be inserted into such operations/calls, as well as optimizing the code for execution during runtime. As a specific illustrative example, binary code (compiled code) may be dynamically optimized during runtime. Here, the program code may include dynamic optimization code, binary code, or a combination thereof.
Similar to a compiler, a translator, such as a binary translator, translates code statically or dynamically to optimize and/or translate the code. Thus, reference to execution of code, application code, program code, or other software environment may refer to: (1) executing a compiler(s) program, optimizing a code optimizer, or translator for compiling program code, maintaining software structure, performing other operations, optimizing code, or translating code, either dynamically or statically; (2) executing main program code including operations/calls, e.g., application code that has been optimized/compiled; (3) executing other program code (e.g., libraries) associated with the main program code to maintain the software structure, to perform operations related to other software, or to optimize the code; or (4) combinations of the above.
Referring now to FIG. 20, a block diagram of an embodiment of a multicore processor is shown. As shown in the embodiment of fig. 20, processor 2000 includes a plurality of domains. Specifically, core domain 2030 includes a plurality of cores 2030A-2030N, graphics domain 2060 includes one or more graphics engines including media engine 2065, and system agent domain 2010.
In various embodiments, system agent domain 2010 handles power control events and power management such that independent units (e.g., cores and/or graphics engines) of domains 2030 and 2060 are independently controllable to dynamically operate at an appropriate power mode/level (e.g., active, turbo charged, sleep, hibernate, deep sleep or other high level configuration power interface, etc. states) depending on the activity (or inactivity) occurring in a given unit. Each of domains 2030 and 2060 may operate at different voltages and/or powers and, furthermore, individual cells within a domain may all operate at independent frequencies and voltages. It should be noted that although only three domains are shown, it should be understood that the scope of the present invention is not so limited, and that additional domains may be present in other embodiments.
As shown, each core 2030 includes, in addition to various execution units and additional processing elements, a low level cache. Here, the various cores are coupled to each other and to a shared cache memory; the shared cache memory is formed by a plurality of units or segments of Last Level Cache (LLC) 2040A-2040N; these LLCs typically include memory and cache controller functionality, and are shared between cores, and possibly graphics engines.
As shown, the ring interconnect 2050 couples the cores together and provides interconnection between the core domain 2030, the graphics domain 2060, and the system agent circuit 2010 via a plurality of ring stops (ring stops) 2052A-2052N, each ring stop being coupled between a core and an LLC segment. As shown in FIG. 20, interconnect 2050 is used to carry various information including address information, data information, acknowledgement information, and snoop/invalidate information. Although ring interconnects are shown, any known on-die interconnect or structure may be utilized. As an illustrative example, some of the above-discussed architectures (e.g., another on-die interconnect, an on-chip system fabric (OSF), an Advanced Microcontroller Bus Architecture (AMBA) interconnect, a multi-dimensional grid architecture, or other known interconnect architectures) may be utilized in a similar manner.
As further depicted, the system agent domain 2010 includes a display engine 2012 that provides control and interfaces to an associated display. The system agent domain 2010 may include other elements such as: an integrated memory controller 2020 to provide an interface to system memory (e.g., DRAM implemented over multiple DIMMs); coherency logic 2022 for performing memory coherency operations. There may be multiple interfaces to enable interconnection between the processor and other circuitry. For example, in one embodiment, at least one Direct Media Interface (DMI)2016 interface and one or more PCIe interfaces are providedTMAn interface 2014. Display engines and these interfaces are typically via PCIeTMThe bridge 2018 is coupled to a memory. In addition, one or more other interfaces may be provided in order to provide communication between other agents (such as additional processors or other circuitry).
Referring now to FIG. 21, shown is a block diagram of a representative core; specifically, a logic block at the back end of a core, e.g., core 2030 from fig. 20. Generally, the architecture shown in fig. 21 includes an out-of-order processor having a front end unit 2170 for fetching incoming instructions, performing various processing (e.g., caching, decoding, branch prediction, etc.), and passing instructions/operations along an out-of-order (OOO) engine 2180. OOO engine 2180 performs further processing on the decoded instructions.
Specifically, in the embodiment of FIG. 21, the out-of-order engine 2180 includes: an allocation unit 2182 for receiving decoded instructions from front end unit 2170 may be in the form of one or more microinstructions or microinstructions and allocating them to the appropriate resources, e.g., registers, etc. Next, instructions are provided to the reservation station 2184, and the reservation station 2184 reserves resources and schedules them for execution on one of the execution units 2186A-2186N. Various types of execution units may exist, including, for example, Arithmetic Logic Units (ALUs), load and store units, Vector Processing Units (VPUs), floating point execution units, and so forth. The results from these different execution units are provided to a reorder buffer (ROB)2188, which takes out the out-of-order results and returns them to correct the program order.
Still referring to fig. 21, it should be noted that both front end unit 2170 and out of order engine 2180 are coupled to different levels of the memory hierarchy. Specifically shown is an instruction level cache 2172, which is in turn coupled to a mid-level cache 2176, which mid-level cache 2176 is in turn coupled to a last level cache 2195. In one embodiment, last level cache 2195 is implemented in on-chip (sometimes referred to as uncore) unit 2190. By way of example, the unit 2190 is similar to the system agent 2010 of fig. 20. As described above, the uncore 2190 is in communication with the system memory 2199, which in the illustrated embodiment is implemented via ED RAM. It should also be noted that the various execution units 2186 within the out-of-order engine 2180 are in communication with a first level cache 2174, which first level cache 2174 is also in communication with a mid-level cache 2176. It should also be noted that additional cores 2130N-2-2130N may be coupled to LLC 2195. While the embodiment of fig. 21 is shown at a high level, it should be understood that various modifications and additional components may exist.
Turning to FIG. 22, a block diagram of an exemplary computer system formed using a processor including an execution unit for executing instructions, in which one or more of the interconnects implementing one or more features are shown according to one embodiment of the present invention. In accordance with the present invention, for example in the embodiments described herein, the system 2200 includes a component such as a processor 2202 for using an execution unit that includes logic for executing an algorithm for processing data. System 2200 shows a PENTIUM III-based systemTM、PENTIUM 4TM、XeonTM、Itanium、XScaleTMAnd/or StrongARMTMA microprocessor, but other systems (including PCs with other microprocessors, engineering workstations, set-top boxes, etc.) may also be used. In one embodiment, sample system 2200 executes WINDOWS available from MICROSOFT CORPORATION of Redmond, WashingtonTMVersions of the operating system, but other operating systems (e.g., UNIX and Linux), embedded software, and/or graphical user interfaces may also be used. Thus, embodiments of the invention are not limited to any specific combination of hardware circuitry and software.
Embodiments are not limited to computer systems. Alternative embodiments of the present invention may be used in other devices such as handheld devices and embedded applications. Some examples of handheld devices include cellular telephones, internet protocol devices, digital cameras, Personal Digital Assistants (PDAs), and handheld PCs. The embedded application may include a microcontroller, a Digital Signal Processor (DSP), a system on a chip, a network computer (NetPC), a set-top box, a hub, a Wide Area Network (WAN) switch, or any other system that may execute one or more instructions in accordance with at least one embodiment.
In the illustrated embodiment, the processor 2202 includes one or more execution units 2208 to implement an algorithm for executing at least one instruction. One embodiment may be described in the context of a single processor desktop or server system, but alternative embodiments may be included in a multi-processor system. System 2200 is an example of a "central" system architecture. The computer system 2200 includes a processor 2202 for processing data signals. As one illustrative example, the processor 2202 comprises a Complex Instruction Set Computer (CISC) microprocessor, a Reduced Instruction Set Computing (RISC) microprocessor, a Very Long Instruction Word (VLIW) microprocessor, a processor implementing a combination of instruction sets, or any other processor device (e.g., a digital signal processor). The processor 2202 is coupled to a processor bus 2210, the processor bus 2210 transferring data signals between the processor 2202 and other components in the system 2200. Elements of system 2200 (e.g., graphics accelerator 2212, memory controller hub 2216, memory 2220, I/O controller hub 2224, wireless transceiver 2226, flash BIOS 2228, network controller 2234, audio controller 2236, serial expansion port 2238, I/O controller 2240, etc.) perform conventional functions well known to those skilled in the art.
In one embodiment, the processor 2202 includes a level 1(L1) internal cache memory 2204. Depending on the architecture, the processor 2202 may have a single internal cache or multiple levels of internal cache. Other embodiments include a combination of both internal and external caches depending on the particular implementation and needs. The register file 2206 is used to store different types of data in various registers, including integer registers, floating point registers, vector registers, packed registers (banked registers), shadow registers, checkpoint registers, status registers, and instruction pointer registers.
An execution unit 2208, including logic for performing integer and floating point operations, also resides in the processor 2202. In one embodiment, the processor 2202 includes a microcode (ucode) ROM to store microcode that, when executed, is used to perform algorithms for some macro-instructions or to process complex scenarios. Here, the microcode may be updatable to handle logical errors/fixes for the processor 2202. For one embodiment, execution unit 2208 includes logic to process packed instruction set 2209. By including the compressed instruction set 2209 in the instruction set of the general purpose processor 2202 and the associated circuitry for executing the instructions, the compressed data in the general purpose processor 2202 can be used to perform operations used by multiple multimedia applications. Thus, multiple multimedia applications are accelerated and executed more efficiently by using the full width of the processor data bus to perform operations on compressed data. This may eliminate the need to transfer a smaller unit of data across the processor's data bus to perform one or more operations (one data element at a time).
Alternative embodiments of execution unit 2208 may also be used in microcontrollers, embedded processors, graphics devices, DSPs, and other types of logic circuitry. The system 2200 includes a memory 2220. Memory 2220 includes a Dynamic Random Access Memory (DRAM) device, a Static Random Access Memory (SRAM) device, a flash memory device, or other memory devices. The memory 2220 stores instructions and/or data represented by data signals that are to be executed by the processor 2202.
It should be noted that one or more of the above-described features or aspects of the invention may be utilized on one or more of the interconnects shown in fig. 22. For example, an on-die interconnect (ODI), not shown, for coupling the internal elements of the processor 2202, implements one or more aspects of the present invention described above. Alternatively, the present invention is associated with a processor bus 2210 (e.g., other known high performance computing interconnect), a high bandwidth memory path 2218 to memory 2220, a point-to-point link to graphics accelerator 2212 (e.g., fabric-compliant peripheral component interconnect (PCIe)), a controller hub interconnect 2222, I/O or other interconnect (e.g., USB, PCI, PCIe) to couple the other illustrated components. Some examples of these components include an audio controller 2236, a firmware hub (flash BIOS)2228, a wireless transceiver 2226, a data store 2224, a legacy I/O controller 2210 containing user input and a keyboard interface 2242, a serial expansion port 2238 (such as a Universal Serial Bus (USB)), and a network controller 2234. Data storage device 2224 may include a hard disk drive, a floppy disk drive, a CD-ROM device, a flash memory device, or other mass storage device.
Referring now to fig. 23, shown is a block diagram of a second system 2300 in accordance with an embodiment of the present invention. As shown in fig. 23, multiprocessor system 2300 is a point-to-point interconnect system, and includes a first processor 2370 and a second processor 2380 coupled via a point-to-point interconnect 2350. Each of the processor 2370 and the processor 2380 may be some version of the processor. In one embodiment, 2352 and 2354 are part of a serial, point-to-point coherent interconnect fabric, e.g., a high performance architecture. Thus, the present invention may be implemented within the QPI architecture.
While shown with only two processors 2370, 2380, it is to be understood that the scope of the present invention is not limited in this regard. In other embodiments, one or more additional processors may be present in a given processor.
A shared cache (not shown) may be included in either processor or external to both processors; but still connected with the processors via the P-P interconnect such that if the processors are placed in a low power mode, local cache information for either or both processors may be stored in the shared cache.
As shown in fig. 23, various I/O devices 2314 are coupled to first bus 2316 and to bus bridge 2318, which couples first bus 2316 to second bus 2320. In one embodiment, second bus 2320 includes a Low Pin Count (LPC) bus. Various devices are coupled to second bus 2320 including, for example, a keyboard and/or mouse 2322, communication devices 2327 and a storage unit 2328 (e.g., a disk drive or other mass storage device that typically includes instructions/code and data 2330). Further, an audio I/O2324 is shown coupled to second bus 2320. It should be noted that other architectures are possible, including variations in component and interconnect architectures. For example, instead of the point-to-point architecture of fig. 23, a system may implement a multi-drop bus or other such architecture.
Turning next to FIG. 24, an embodiment of a system on a chip (SOC) design in accordance with the invention is depicted. As a particular illustrative example, SOC 2400 is included in a User Equipment (UE). In one embodiment, a UE refers to any device used by an end user for communication, such as a handheld phone, a smart phone, a tablet computer, an ultra-thin notebook, a notebook with a broadband adapter, or any other similar communication device. Typically, the UE is connected to a base station or node, which may essentially correspond to a Mobile Station (MS) in a GSM network.
Here, SOC 2400 includes two cores — 2406 and 2407. Similar to the discussion above, cores 2406 and 2407 may conform to an instruction set architecture (e.g., based onArchitecture CoreTMAdvanced micro device corporation (AMD) processors, MLP-based processors, ARM-based processor designs, or their customers), and their licensees or users. Cores 2406 and 2407 are coupled to cache control 2408 associated with bus interface unit 2409 and L2 cache 2411 for communicating with other portions of system 2400. Interconnect 2410 includes an on-chip interconnect (such as an IOSF, AMBA, or other interconnect discussed above), which may implement one or more aspects described herein.
In addition, the system shows peripherals used for communication, such as a bluetooth module 2470, a 3G modem 2475, GPS 2485, and WiFi 2485. As described above, the UE includes a radio for communication. Therefore, not all of these peripheral communication modules are required. However, in some forms of UE, a radio for external communication will be included.
Fig. 25-27 disclose examples of data encoding without a dedicated STREAM path. In embodiments of the interface described herein, a "STREAM" path may be provided to differentiate betweenOn-chip interconnect (IDI) services andsystem on chip architecture (IOSF) services, both provided at the link layer. Link Control Packets (LCPs) on the PHY layer may also need to be marked. In one embodiment, one flow path is provided in each cluster of 20 data paths.
However, in some embodiments, an interface may be provided without a dedicated STREAM path. For example, STREAM path data may be provided within the data path during idle periods in order to provide equivalent functionality. The elimination of flow paths saves about 5% area because one flow path can be provided in every 20 data paths.
In one embodiment, during the pre-data time, 20 data lanes from the middle track may be placed high to indicate one data type and 20 data lanes may be placed low to indicate a second data typeThe class (e.g.,system on chip architecture (IOSF)). To represent additional data categories, such as Link Control Packets (LCPs), the lanes may be divided into two or more groups, and individual bits may be encoded into each group. The LCP may also be coded into a post-data time, for example, by stopping flit traffic and controlling the "VALID" path from the middle rail to be 0 or 1.
There are a variety of ways in which this can be accomplished. In fig. 25, an arrow 2510 marks an example pre-data symbol time, and an arrow 2520 marks an example post-data symbol time. In the example of fig. 25, in addition to data lanes, which may be provided in groups of n lanes each (where n ═ 20 in one example), a "STROBE" lane may be provided, and each group may include a "VALID" lane. The "STREAM" path is also disclosed by way of example to illustrate that signaling may be required without the teachings of the present specification. With the method of fig. 26 and 27, the STREAM path may be eliminated, providing a space savings of approximately 5% in one embodiment.
During the pre-data period, for example, the data path of the middle track in a triple data (trideata) scheme remains idle. However, they can be driven to 0 or 1 without loss of electrical integrity. Thus, the free data path is useful for encoding stream data without a separate flow path.
In one embodiment, the IOSF and IDI need only be distinguished in the pre-data period. Thus, all the channels can be driven high or low to represent one of two possibilities. However, the present specification is not limited thereto. It is also possible to provide k bits of stream data by dividing the stream into k groups and driving values onto each group to represent 2kAnd (4) grouping. The size of the groups may be uniform, but this is not essential. If no vias are needed, they may be idle in the mid-track state.
It is also possible to use only k lanes instead of split groups. However, the use of packets provides redundancy that is beneficial to embodiments in which reliability is superior to various other embodiments.
Post data may also be provided for encoding the LCP packet. This is useful because in the case where only two types of packets (e.g., IOSF and IDI) are provided by the link layer, the datapath can receive a uniform value while the LCP is encoded separately. Encoding the LCP post-data is possible because, in some embodiments, LCP always takes precedence over flits. When a proxy needs to send LCP over PHY, it can terminate flit traffic by providing backpressure on the link layer.
Similar to the manner in which multiple link layer categories may be provided as described above, different types of LCPs may be encoded onto the lanes during the symbol time of the following data period. For example, "00" on lanes 0 and 1 may be requested to enter hardware re-centering. This may occur when a Cyclic Redundancy Check (CRC) error is encountered at a higher rate than expected, so that one of the agents determines that the link needs to be retrained so that the clock can be properly "centered". In an example, "01" may be an ACK for the request. These are provided by way of non-limiting example only, and it should be noted that many different LCP requests and responses may be usefully encoded in the scheme. Furthermore, encoding the same value over multiple lanes, as described above, may help reduce errors. In an embodiment, holding the via at the middle rail indicates that no LCP is sent.
Furthermore, in some embodiments, it may be desirable to send the LCP in quiet hours after the post-data period. To do so, for example, the VALID path may be pulled to 0 just before the LCP is sent. During quiet periods, the VALID path may remain at the middle rail everywhere else.
In some embodiments, STROBE, VALID, and STREAMID for all clusters may be driven the same as each other. In other embodiments, they may be driven individually to provide enhanced capabilities, e.g., re-centering only individual clusters, while other clusters continue to receive normal traffic.
Fig. 26 is a flow chart illustrating a method 2600 of providing embedded flow path data. In the example of fig. 26, pre-data encoding is used.
In block 2610, the stream path encoder encodes a stream path identifier or a class identifier to identify a class of data to track.
In block 2620, the lane driver drives the encoded species identifier onto the data lane, still during the pre-data time shown in fig. 25. In some embodiments, where a multi-bit flow path identifier is required, this may include dividing the data paths into an appropriate number of groups. For example, if there are 20 lanes and four bits are required to represent up to 16 data classes, the data lanes may be divided into four groups of five lanes each. In each group, all five lanes will be driven to the same state. In case there is an additional lane that cannot receive a value, then this additional lane may be left at the intermediate rail.
In block 2630, a pre-data period expires. The data path must now be freed for substantive data use.
In block 2640, the lane driver drives substantive data onto the data lane.
In block 2690, the method is complete.
Figure 27 is a flow diagram illustrating a method 2700 of providing additional post-dataflow information (e.g., LCP signals).
In block 2710, the flow path encoder encodes a flow path identifier or a category identifier to identify a category of data to track.
In block 2720, the lane driver drives the encoded species identifier onto the data lane, still during the pre-data time shown in fig. 25. In some embodiments, where a multi-bit flow path identifier is required, this may include dividing the data paths into an appropriate number of groups. For example, if there are 20 lanes and four bits are required to represent up to 16 data classes, the data lanes may be divided into four groups of five lanes each. In each group, all five lanes will be driven to the same state. In case there is an additional lane that cannot receive a value, then this additional lane may be left at the intermediate rail.
In block 2730, the pre-data period expires. The data path must now be freed for substantive data use.
In block 2740, the lane driver drives substantive data onto the data lane.
In block 2750, the data cycle ends. The data path may now again be used as the flow identifier.
In block 2760, for example, agent a determines that it is needed to provide an LCP. Thus, agent a stops sending flits.
In block 2770, the flow encoder encodes the LCP. For example, a code for application "RECENTER" may be provided, and another code for agent B (FIG. 3) may be provided to provide an "ACK" for the request. Additional LCP codes are also available.
In block 2780, the lane driver drives the LCP code onto the datapath. The two agents then perform the requested action.
In block 2790, the method is complete.
Fig. 28-30 illustrate one embodiment of the centering state shown in fig. 17. In this embodiment, when the interface is brought up from a powered down state, the link must be "trained" to run at full speed. Training is necessary because values may "drift" over time and between different conditions (e.g., temperature). Particularly with high speed links like PCIe, it is necessary to properly center the signal in both the time (horizontal) and voltage (vertical) domains to avoid bit errors that can be caused by parameter drift.
Training may include "centering" or "re-centering" the clock signal to ensure that the signal is properly centered in both the time domain and the voltage domain. At any given speed, the clock has a window in which the samples of the datapath will produce the correct value. Samples outside the window may produce bit errors. The faster the clock, the smaller the window. With added complexity, there are separate windows for vertical Vref centering and horizontal phase centering. Thus, in one example, and particularly in high speed interconnects with narrow windows, an important aspect of training is adjusting both Vref and the phase of the clock so that data is sampled at or near the center of the window to minimize bit errors.
In one example, training includes supplying an interface with a known value, such as PRBS, and then sampling back a bit pattern. After sampling, the clock may be shifted vertically in the voltage domain (Vref) or horizontally in the time domain (phase). The interface detects when bit errors begin to occur, thereby constructing a two-dimensional "eye," which may be substantially diamond-shaped. In short, values inside the eye cause no (or little, acceptable) bit errors, while values outside the eye cause an unacceptable number of bit errors.
Centering may include a combination of both hardware and software. In some embodiments, software centering may be run from a Basic Input Output System (BIOS) chip. In one example, hardware for basic or initial centering operations is provided, which may provide a coarse centering solution. Software may then be used to supplement and improve the initial centering. Software centering provides additional flexibility and, in some cases, higher logic programmability than hardware centering. This may solve problems such as, for example, hardware centering, finding that it considers the content to be "near center" for the eye, when a closer center may actually be obtained via additional steps/passes.
In one example, the hardware includes 32 quantized delay settings for phase adjustment, but any number of quantized delay settings may be provided. Vre fSettings may be similarly quantified.
In a first (hardware-only) stage, centering may include a sweep of quantized phase settings across each pass, taken at a nominal voltage V (e.g., the original value or the last calculated value). After each scan, the error detector lists the number of errors (if any) encountered in each step/pass.
Error detection in this embodiment is an example of transmitter (Tx) adaptation, i.e., the Tx value is adjusted based on the remote receiver (Rx) metric available for Tx on the reverse channel. Software running on the core, typically on the upstream die, may use the sidebands (see fig. 6) to access the local die and the remote die, which act as back channels. In HPI, this adaptation occurs in both directions simultaneously, since the reverse channel is not on the same wire as the link layer.
In an embodiment, each scan is aggregated. Thus, rather than scoring the number of errors on any one pass, the scan scores the number of errors across all passes. This means that in some cases the arrangement that is best for the independent vias may not be best for the interconnect as a whole. Such a phase sweep may be performed in hardware at initial start-up and may be used to perform coarse centering, where an approximate center phase is selectedIn an embodiment, the phase scanning is performed at full speed such thatIs suitable for full speed operation. In other embodiments, the initial hardware scan may be performed first at partial speed (e.g., at half speed), and then refined at full speed.
After the phase scan is complete, the interface is placed at full speed (if needed), and software instructions may be loaded from the BIOS to continue to improve centering in software. In this second phase (which is performed in software, in one example), the voltage (V)ref) Values are similarly swept through the quantization settings. Each scan may be as followsIs performed. In other words, each path is utilized inPhase setting of (a) performs (V)refScanning to determine the optimum V at that phaserefThe value is obtained. The offset σ is then incremented (e.g., by 1) and is atEither side of (e.g.,) Upper VrefIs scanned. If the eye is not fully defined, σ can be increased again and the process can be repeated until the eye is fully defined (see FIG. 30), which occurs when V is fully definedmax≈Vmin(wherein, VmaxIs selected atHas a maximum phase with an acceptable error rate, and VminIs selected inWith a minimum phase having an acceptable error rate).
As with the phase scan, each voltage scan results in zero or more errors, and at each VrefThe results at the value are aggregated across all lanes. In an embodiment, V is performed at an operational speed (8 Ghz in one example)refScanning to ensure that the results are effective for operational speed.
In a third stage, the software may construct a two-dimensional "eye" that includes a center point based on the values identified in the first two stages. The eye may be a diamond or other suitable shape. Determining the optimal center of the eye may be meaningful because the eye opening may be asymmetric, non-monotonic around the dominant inflection point, and somewhat time-varying. Thus, eye centering may be performed in software. In one example, centering includes findingAndand the value of the midpoint between the two, and VmaxAnd VminAnd the value of the midpoint between the two. From the midpoint, the software can calculate a suggested center
The software may then select four inflection points along the diamond edges, and at the inflection points, four vertices along the diamond (at V)max、Vmin、Andat) run test mode. If the eye has been constructed and properly centered, these eight points should all fall well within the eye and therefore should have an acceptable error rate. If one or more of the eight points results in an unacceptable error rate, the eye may need to be adjusted to improve performance. For example, if an inflection point fails, the boundary of the eye may need to be adjusted one quantum inward in one or both axes. A similar test then needs to be performed on the new value to ensure that it passes. When the point is adjusted accordingly, the centerAnd may also be adjusted accordingly.
The error detection described in one or more of the above-described stages may be enhanced by providing crosstalk patterns as shown in fig. 29A and 29B. In this example, it can be seen that the testing of one path includes: the path under test is selected as the "victim" path. A test pattern (including, for example, PRBS) may be driven on the path. Adjacent vias and at most l vias at far away (in this example, l ═ 4) are "aggressor" vias. These paths receive the crosstalk pattern, which in this case is a logical inverse of the test pattern, thereby approximating a "worst case" scenario. The remaining path is the cube. This test mode is particularly helpful in centering the eye, as it is important at this stage to detect edge conditions and ensure that they are reasonably within the eye.
Fig. 28 provides a centering flow diagram as described in this specification.
In block 2810, the interconnect wakes up, e.g., starts up after the machine has been shut down.
In block 2820, the interconnect performs initial voltage centering as described above. This may include performing a voltage sweep in hardware and selecting the setting that results in the least total bit error.
In block 2840, once the coarse voltage centering is done in hardware (if the voltage centering is done at a reduced speed (e.g., at half speed)), the interface may be placed at its full speed.
In block 2850, phase centering is performed (e.g., in software). In some examples, the software may also improve voltage centering as described herein. As described above, phase centering in one embodiment includes sweeping each voltage V0Phase of + -sigmaIn other words, a voltage is selected and the phase is scanned with this voltage.
In block 2870, the eye is constructed in software.
In block 2880, eye centering is performed. As described above, in one embodiment, this includes selecting four vertices and four inflection points, and stress testing each point to determine if it passes. If a point fails, the vertex or inflection point may be adjusted inward and tested again. Center of eyeAnd may also be adjusted accordingly.
29A and 29B illustrate an example victim/aggressor test mode as described above. In this example, up to four vias above or below the victim (tested) via are filled with aggressor patterns. The remaining path is the cube. In an embodiment, the aggressor mode is an exact bitwise inversion of the PRBS test mode on the victim way.
Fig. 30 is an illustration of an example of an eye 3000. In this illustration, the eye 3000 is suitably diamond shaped for ease of illustration of some important features. However, the features of the eye 3000 are not required and typically will not be normal or monotonic. The diamond shape defined by eye 3000 may be left, right, up or down slanted, and the edges need not consist of perfectly collinear lines. This is particularly true once the eye 3000 is centered.
In this example, the eye 3000 initially scans each V across multiple phase quantarefSelection 3010 to define to identify the VrefTwo selected phase endpoints 3020. The scanning may be from V03034 and proceeds in the quantization steps above and below, up to Vmax3030 and V min3032 is identified. In the case of the present example,3042 and3040 all appear at V0A line, but in practice one or both may be present on some other line and need not be present on the same line.
An inflection point 3050 is also selected. In this example, V0To VmaxAn odd number of quanta is included so that inflection points 3050-1 and 3050-2 can be selected at the "true" center point between the two. However, V0To VminAn even number of quanta is included so that the inflection points 3050-3 and 3050-4 must be selected off-center. These may be selected by floor functions, ceiling functions, or on either side of the interval depending on design choice.
Initial center may also be selectedAgain, since these values are quantized, this may not be the geometric center, but may be off-centered in any direction by one or more quanta. During centering, Vmax 3030、Vmin 3032、3042、3040, and an inflection point 3050 are used as test points. As a result of the testing, one or more of these dots may be adjusted up or down, left or right by one or more quanta, thereby making the shape of eye 3000 more irregular. When the test points are adjusted, the centerAnd can be adjusted correspondingly according to the needs. Once centering is complete, all paths are set toTo ensure optimum operation.
Fig. 31 is a flow diagram of an eye centering method 3100 in accordance with one or more examples of the present description. Some embodiments of the method 3100 may be performed in software, but the method may also be performed in hardware, or in a combination of hardware and software.
In block 3110, software selects a test point, e.g., V of FIG. 30max 3030、Vmin 3032、3042、3040 and an inflection point 3050.
In block 3120, the software performs a stress test at each test point on each pathway using the victim/aggressor pathway described in fig. 29A and 29B.
In decision block 3130, ifAll points pass, then in block 3150, the phase and voltage of the center of the eye may be utilizedAll the vias are provided.
In block 3140, if not all points pass, the failed points may be adjusted inward. Control then passes to block 3120 to test for new point(s). This continues until all points have passed and the datapath has been appropriately configured in block 3150.
In block 3190, the method is complete.
In some examples, an operation known as "re-centering" may also occur as desired. For example, re-centering may be required when an unacceptable number of errors occur during operation. In one example, the unacceptable error rate is more than about 1012One of the bits has a Cyclic Redundancy Check (CRC) error. When this occurs, the flit is stopped and phits takes over the interconnect to perform the phase scan in hardware (since the machine is running, software may not be available and therefore cannot load the BIOS).
Centering may take several microseconds, which is unacceptably slow for periodic centering in high speed interconnects. Thus, in one embodiment, re-centering occurs only when an error tracker in the interface link layer detects that the number of CRC errors received in a time span is greater than a threshold. For example, every 1012A one bit CRC error may be acceptable. But more likely is an indication of a true error. Thus, the receiving agent may send a side message informing the transmitting agent that re-centering is required. When this occurs, an agent acknowledgement message is sent and both agents enter into re-centering.
The messages on the sideband may be Link Control Packets (LCPs), which are special classes of packets sent on the interface sideband. The sidebands may operate at speeds on the order of magnitude slower than the link layer. For example, the interface sideband may operate at 100 to 400 MHz.
Some interface designs may choose to use signaling over the data path for low frequency or continuous CDR tracking, which consists of flits/packets scrambled with PRBS 23. During quiet periods, when there are not enough packets on the datapath, the PHY inserts a virtual LCP so that this tracking can be maintained.
Turning now to fig. 32-34, in some embodiments it may be beneficial to design the system in stages by first designing low-order blocks, and then aggregating the low-order blocks into functional elements. An Intellectual Property (IP) block (or "computing block") may refer to a reusable element of logic, circuitry, or chip layout. For example, the SoC may include multiple IP blocks, including multiple instances of the same IP block. In some instances, the IP block may support multiple functions, in some cases, implemented by one or more devices included in the IP block. The system may be designed to implement by defining a set of IP blocks. For example, one team may design a low-level functional block and make that block available to other teams, which acts as a drop-in component for subsequent design. In this type of multi-layer design process, it is beneficial to provide some functionality as "hard IP blocks" or "hard IPs" (meaning that IP blocks have fixed routes that may not change during higher level design). The hard IP block may include, for example, analog components that are sensitive to routing changes. For some analog components, rerouting may result in changes in impedance or capacitive reactance that may affect the function of some sensitive analog components. Thus, a development team responsible for designing hard IP blocks can optimize these parameters and ensure that each of the components operates within acceptable parameters. Once the route is fixed, it should not be changed by downstream designers.
Other IP blocks may be provided as "soft IP". Soft IP may include, for example, digital components that are less sensitive to routing variations. Thus, the soft IP blocks may have fixed logic functions, but may be arranged according to the design requirements of a particular design. For example, the PHY layer may have a physical PHY implemented in hard IP, and a logical PHY implemented in soft IP.
Although it is advantageous to provide some types of functional blocks in hard IP, the limitation of hard IP is that the block is inflexible by design and definition. Thus, in one example, an SOC designer may require 50 data paths. The designer may be working with a hard IP block that provides 20 data paths per cell. Because the routing of the hard IP blocks is fixed at design time, the SOC designer must use an integer number of hard IP blocks. In particular, p hard IP blocks are used, where each block provides n data lanes, and all h n p total hard IP data lanes are provided. In the case where the system design requires a total of k data lanes, k ≠ h may be encountered. For example, when k is 50 and n is 20, p ═ k/n ═ 3 hard IP blocks require h ═ n ═ p ═ 60 hard IP datapaths available. This result is acceptable as long as h ≧ k, or in other words, as long as p hard IP blocks provide at least enough data paths. During the layout of the final SOC, the optimization software will not connect the additional 10 data paths to any power supply or logic gates, so that the only cost associated with having additional data paths is a certain amount of additional space on the silicon wafer. Since hard IP blocks tend to be small compared to the entire wafer, the wasted space is not excessive. To facilitate the use of hard IP blocks, each data path may have a separate enable, such that unused data paths may be ignored. It is also noted that other combinations of vias from clusters may be used depending on the circumstances (e.g., unused vias may be distributed in the middle of a cluster to reduce crosstalk).
Fig. 32 discloses a block diagram of an example die, showing that a logical PHY may be implemented in soft ip (sip), while a physical PHY may be implemented in hard ip (hip). In this example, an internal standard interface (OPIS, On Package Interface (OPI) PHY interface specification) is defined to facilitate partitioning. A link PHY interface (LP I) is also provided.
The partition allows configuration aspects such as path reversal to move to soft IP. Detailed in dashed box) is constructed as a cluster that allows flexibility (e.g., port width) without sacrificing performance (e.g., speed).
Fig. 33 is an illustrative example of placement of hard and soft IPs in accordance with one or more examples of the present specification. In fig. 33, an upper layer 3310 provides a higher level of logic and may be implemented in any suitable form. Soft IP3320 provides the logical PHY. Several hard IP blocks 3330 are also provided. Specifically, hard IP blocks 3330-1, 3330-2, and 3330-3 each have 20 data paths (n-20) and have necessary support paths. In this embodiment, the PHY has all 50 lanes (k 50). Therefore, three hard IP blocks 3330 are used because all 60 available paths in hard IP (h ═ 60). Since the hard IP block is fixed, 10 additional lanes are included. In this example, each hard IP block 3330 may provide separate enablement for each data path. Accordingly, lanes 10-19 on hard IP block 3330-3 are disabled. During layout, they will receive no route, no power supply, no clock, and no other signals. The vias 10-19 are shown as being disabled by way of example only. In other embodiments, any 10 vias may be disabled to achieve the same effect, depending on design requirements.
Fig. 34 illustrates a problem that may occur with the use of hard IP blocks and soft IP blocks (i.e., path inversion). This may be illustrated with an example PHY package that provides both a transmitter and a receiver. In this example, the transmitter and receiver may each be provided on separate SOC dies and then integrated onto a single package. Each die may include, for example, a physical PHY provided in hard IP, and a logical PHY provided in soft IP. Because hard IPs, soft IPs, and/or dies may be provided by different groups within the same vendor or by different vendors, in some cases, the data lanes of one group may be reversed from the data lanes of another group. In the case of an SOC designer, the designer may not know in advance how the OEM intends to implement the design, and whether the vias are reversed.
Thus, lane reversals may be encountered at various interfaces. In one case, a path reversal may occur between the hard IP and the soft IP. In other cases, via inversion may occur between two dies on the same package.
To handle such lane inversion, "bowtie" may be used to route the data lanes to each other. However, since these data paths must cross each other, at least 2 additional layers need to route (either on the die or on the package) the bow tie. This introduces additional design complexity
Fig. 34 includes three packages 3410. For example only, each encapsulation 3410 includes hard IP 3330, soft IP3320, and upper layer logic 3310. As previously described, hard IP 3330 provides the physical PHY, soft IP3320 provides the logical PHY functions, and upper layers 3310 provide additional upper layer logic.
For example, in package 3410-1, die 1(Tx) is a PHY transmitter and die 2(Rx) is a PHY receiver. As shown in this example, the data path of die 1 is the path inverted from the data path of die 2. Thus, if a bow tie is used, additional layers may be required so that the connections can cross over each other. This increases the complexity of the design.
In package 3410-2, there is a path reversal between hard IP 3330 and soft IP3320 of die 1. In this case, a bow tie connection must be formed on the die 1. Again, the bow tie would require additional layers on the die 1 so that the vias can be properly matched. As previously mentioned, this results in additional design complexity.
However, in the example of packages 3410-3, via reversals may be repaired and vias properly aligned without requiring bow tie connections and the additional complexity associated therewith. In particular, in package 3410-3, soft IP3320 includes multiplexer-like programmable switches 3480. The switch 3480 may not be a conventional multiplexer having a selection signal for selecting between inputs, but the switch 3480 may be similar in internal structure to a multiplexer, with the number of inputs being the same as the number of outputs in the present example having a difference.
The programmable aspects of switch 3480 include a switch that includes at least two modes, one mode for routing directly through a via and another mode for routing a via in reverse order. Other programming options may also be provided as desired. To this end, the data path may have a virtual fuse characteristic. At start-up, a fuse pull (fuse pull) device may determine whether to reverse the path on package 3410-3. If the path is to be reversed, programmable switch 3480 will be blown to perform the path reversal. If the path is not to be reversed, the fuse drawer blows the multiplexer 3420 so as not to reverse the path. With known orientations at design time, the designer of package 3410-3 can program the fuse clips to provide the correct orientation.
Advantageously, this architecture takes into account the fact that the designer of soft IP3320 may not know the direction of the path in advance. For example, packages 3410-3 may be provided to an Original Equipment Manufacturer (OEM) who may orient packages 3410-3 according to its own design requirements. In this case, the OEM may know the orientation and may manually set the fuse puller to properly orient switch 3480. In another example, at startup, the packages 3410-3 may perform a training sequence to determine whether the data path is directly through the route or is cross-routed. Based on the results of the training sequence, software can then program the fuse puller to provide the appropriate direction for switch 3480.
While the present invention has been described with respect to a limited number of embodiments, those skilled in the art will appreciate numerous modifications and variations therefrom. It is intended that the appended claims cover all such modifications and variations as fall within the true spirit and scope of this present specification.
A design may go through various stages, from creation to simulation to fabrication. The data representing the design may represent the design in a number of ways. First, as used in simulations, the hardware may be represented using a hardware description language or another functional description language. Furthermore, at some stages of the design process, a circuit level model with logic and/or transistor gates may be generated. Furthermore, most designs, at some stage, reach a level of data representing the physical layout of various devices in the hardware model. Where conventional semiconductor fabrication techniques are used, the data representing the hardware model may be the data specifying the presence or absence of various features on different mask layers for masks used to produce the integrated circuit. In any representation of the design, the data may be stored in any form of a machine-readable medium. A memory or magnetic or optical storage (e.g., a disk) may be the machine-readable medium for storing the information transmitted via modulation or other generation of optical or electrical waves for transmission of such information. When an electrical carrier wave indicating or carrying the code or design is transmitted, to the extent that copying, buffering, or re-transmission of the electrical signal is performed, a new version is produced. Thus, a communication provider or a network provider may store on a tangible, machine-readable medium, at least temporarily, such as a replica of information encoded into a carrier wave, embodying techniques of embodiments of this specification.
A module as used herein refers to any combination of hardware, software, and/or firmware. As an example, a module includes hardware, such as a microcontroller, associated with a non-transitory medium to store code adapted to be executed by the microcontroller. Thus, in one embodiment, a reference module, refers to hardware, which is specifically configured to identify and/or execute code for retention on non-transitory media. Moreover, in another embodiment, the use of modules refers to a non-transitory medium including code, which is particularly suited for implementation by a microcontroller to perform predetermined operations. And as such it may be inferred that in yet another embodiment, the term module (in this example) may refer to a combination of a microcontroller and a non-transitory medium. In general, the boundaries of modules shown as separate typically vary and may overlap. For example, the first and second modules may share hardware, software, firmware, or a combination thereof, while perhaps retaining some separate hardware, software, or firmware. In one embodiment, use of the term logic includes hardware, such as transistors, registers, or other hardware (e.g., programmable logic devices).
In one embodiment, the use of the phrase "configured to" refers to arranging, placing together, manufacturing, offering for sale, importing and/or designing a device, hardware, logic or element to perform a specified or determined task. In this example, a device or element is still "configured" to perform a specified task when it is not operating, if it is designed, coupled, and/or interconnected to perform the specified task. As a purely illustrative example, a logic gate may provide a 0 or a 1 during operation. But the logic gate is "configured to" provide an enable signal to the clock, which excludes every possible logic gate that may provide a 1 or a 0. Instead, the logic gate is a logic gate that is coupled in some manner during operation, with either a 1 or 0 output used to enable the clock. It is again noted that use of the term "configured to" does not require operation, but instead focuses on the latency of the device, hardware, and/or elements in which the latency of the device, hardware, and/or elements is designed to perform a particular task when the device, hardware, and/or elements are operating.
Furthermore, in one embodiment, the use of the phrases "for," "capable of/capable of being used to" and/or "operable to" refer to some means, logic, hardware, and/or element for which the use of the means, logic, hardware, and/or element is enabled in a particular manner. In one embodiment, it is noted that the use of "for," capable of, or "operable for," as described above, refers to the latent state of a device, logic, hardware, and/or element, wherein the device, logic, hardware, and/or element is not operational but is designed in such a way as to enable use of the device in a particular manner.
As used herein, a value includes any known representation of a number, state, logic state, or binary logic state. In general, the use of logic levels, logic values or logic values, also referred to as 1's and 0's, merely represent binary logic states. For example, a 1 refers to a high logic level and a 0 refers to a low logic level. In one embodiment, a memory cell, such as a transistor or flash memory cell, may be capable of holding a single logic value or multiple logic values. However, other representations of values have been used in computer systems. For example, the decimal number ten may also be represented as a binary value of 1010 and a hexadecimal letter a. Thus, a value includes any representation of information that can be maintained in a computer system.
Further, a state may be represented by a value or portions of a value. As an example, a first value, such as a logical one, may represent a default or initial state, while a second value, such as a logical zero, may represent a non-default state. Further, in one embodiment, the terms reset and set refer to default and updated values or states, respectively. For example, the default value may comprise a high logical value (i.e., reset) and the updated value may comprise a low logical value (i.e., set). It should be noted that any combination of values may be used to represent any number of states.
The embodiments of methods, hardware, software, firmware or code described above may be implemented via instructions or code stored on a machine-accessible, machine-readable, computer-accessible, or computer-readable medium that are executable by a processing element. A non-transitory machine-accessible/readable medium includes any mechanism, such as a computer or electronic system, that provides (i.e., stores and/or transmits) information in a form readable by a machine. For example, a non-transitory machine-accessible medium includes: random Access Memory (RAM) (e.g., static RAM (sram) or dynamic RAM (dram)); a ROM; a magnetic or optical storage medium; a flash memory device; an electrical storage device; an optical storage device; an acoustic storage device; other forms of storage devices that retain information received from transitory (propagated) signals (e.g., carrier waves, infrared signals, digital signals); and the like, that are used to distinguish them from non-transitory media from which information may be received.
Instructions for programming logic to perform embodiments of the invention may be stored in memory in the system, such as DRAM, cache, flash, or other storage. Further, the instructions may be distributed via a network or by other computer readable media. Thus, a machine-readable medium may include any mechanism for storing or transmitting information in a form readable by a machine (e.g., a computer), but is not limited to, floppy diskettes, optical disks, compact discs (CD-ROMs), read-only memories (ROMs), and magneto-optical disks, read-only memories (ROMs), Random Access Memories (RAMs), erasable programmable read-only memories (EPROMs), electrically erasable programmable read-only memories (EEPROMs), magnetic or optical cards, flash memories, or tangible, machine-readable storage for use in transmitting information over the internet via electrical, optical, acoustical or other forms of propagated signals (e.g., carrier waves, infrared signals, digital signals, etc.). Thus, a computer-readable medium includes any type of tangible, machine-readable medium suitable for storing or transmitting electronic instructions or information in a form readable by a machine (e.g., a computer).
The following examples relate to embodiments according to the present description. One or more embodiments may provide an apparatus, system, machine-readable storage, machine-readable medium, hardware-based and/or software-based logic, and method for providing a hard Intellectual Property (IP) subcomponent and a soft IP component having at least one soft IP block, wherein the hard Intellectual Property (IP) subcomponent includes p hard IP blocks, wherein each hard IP block is to provide n data lanes and has fixed routing, and wherein the p hard IP blocks are to provide h n p available hard IP lanes; a soft IP component having at least one soft IP block includes routable logic and provides k data lanes, where k ≠ h.
In one or more embodiments, the data lanes have individual enables and k-h lanes are to be disabled.
In one or more embodiments, the soft IP block further comprises a programmable switch for selectively mapping the data paths according to a switching pattern.
In one or more embodiments, the switching pattern includes at least: a direct path mapping mode, in which paths are to be mapped in numerical order; and a cross-mapping mode, in which lanes are to be mapped in reverse numerical order.
In one or more embodiments, the switching mode can be programmed via a fuse puller.
In one or more embodiments, the switching pattern can be programmed via a training sequence.
In one or more embodiments, the fuses can be programmed via read-only memory.
In one or more embodiments, each lane is to receive an uncorrelated pseudo-random bit sequence for centering.
In one or more embodiments, at least some of the hard IP blocks are to receive separate VALID signals.
In one or more embodiments, at least some of the hard IP blocks are used to receive a separate STREAM ID.
In one or more embodiments, an interconnected system, comprising: a first agent; a second agent; and an interconnect to communicatively couple the first agent to the second agent, the interconnect comprising: a hard Intellectual Property (IP) subcomponent including P hard IP blocks and a soft IP component having at least one soft IP block; wherein each hard IP block is configured to provide n data lanes and has a fixed route, and wherein the p hard IP blocks are configured to provide h n p available hard IP lanes; the soft IP component having at least one soft IP block comprises routable logic and provides k data lanes, wherein k ≠ h.
In one or more embodiments, there is also disclosed a method of providing an interconnect, comprising: placing a hard Intellectual Property (IP) subcomponent including p hard IP blocks, placing a soft IP component having at least one soft IP block; wherein each hard IP block is configured to provide n data lanes and has a fixed route, and wherein the p hard IP blocks are configured to provide h n p available hard IP lanes; the soft IP component having at least one soft IP block comprises routable logic and provides k data lanes, wherein k ≠ h.
In one or more embodiments, the method further comprises disabling k-h lanes.
In one or more embodiments, the method further comprises selectively mapping the data lanes according to a switching pattern.
In one or more embodiments, selectively mapping data lanes according to a switching pattern includes: direct path mapping mode in numerical order.
In one or more embodiments, selectively mapping data lanes according to a switching pattern includes: the lanes are interleaved in reverse numerical order.
In one or more embodiments, the method further comprises providing a pseudo-random bit sequence for each lane for centering.
In one or more embodiments, an interconnect apparatus, comprising: a clock generator for generating a clock signal; a phase shifter for phase shifting the clock signal by a test phase; an error rate detector for measuring an error rate of the test phase; and an operable phase selector for selecting an operable phase of the clock signal based at least in part on an error rate of the test phase.
In one or more embodiments, an interconnect apparatus, comprising: a stream path encoder for encoding a class identifier for the data packet; and a lane driver for driving the species identifier onto at least one of the n data lanes during a non-data time of the data packet.
Reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
In the foregoing specification, specific embodiments have been given with reference to specific exemplary embodiments. It will, however, be evident that various modifications and changes may be made thereto without departing from the broader spirit and scope of the invention as set forth in the appended claims. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense. Moreover, the foregoing use of embodiment and other exemplarily language does not necessarily refer to the same embodiment or the same example, but may refer to different and distinct embodiments, as well as potentially the same embodiment.
Claims (9)
1. An integrated circuit die, comprising:
a Hard IP (HIP) subcomponent including a plurality of hard logic blocks, wherein the hard logic blocks are associated with a plurality of data lanes having separate per-lane enablement;
a plurality of protocol specific logic blocks for communicating data in accordance with a corresponding plurality of data communication protocols, the plurality of protocol specific logic blocks comprising:
a first protocol specific logic block for communicating data in accordance with a first data communication protocol; and
a second protocol specific logic block for communicating data in accordance with a second data communication protocol; and
a Soft IP (SIP) component having at least one soft logic block comprising routable logic to selectively map data lanes of the plurality of data lanes to the first protocol specific logic block and to selectively map data lanes of the plurality of data lanes to the second protocol specific logic block.
2. The integrated circuit die of claim 1,
the first protocol specific logic block includes PCIe logic, an
The second protocol specific logic block includes logic to enable communication using either PCIe or a different protocol.
3. The integrated circuit die of claim 1 or 2, further comprising:
a host processor coupled to a first hard logic block of the plurality of hard logic blocks via the first protocol specific logic block and coupled to a second hard logic block of the plurality of hard logic blocks via the second protocol specific logic block.
4. The integrated circuit die of claim 1,
each of the plurality of data lanes represents a different pair of signals, including:
a transmit pair for transmitting data via different signal paths; and
a receive pair for receiving data via different signal paths.
5. The integrated circuit die of claim 3, further comprising:
sideband links, including bidirectional low frequency control signal links, for transmitting and receiving messages containing control messages.
6. The integrated circuit die of claim 3 or 5, further comprising:
a multi-protocol on-die fabric that provides an infrastructure to communicate between the host process, the first protocol specific logic block, and the second protocol specific logic block.
7. The integrated circuit die of any of claims 1-6,
the plurality of data communication protocols includes a memory access protocol and an input/output (I/O) communication protocol.
8. An interconnect device, comprising:
a clock generator for generating a clock signal;
a phase shifter for phase shifting the clock signal by a test phase;
an error rate detector for measuring an error rate of the test phase; and an operable phase selector for selecting an operable phase of the clock signal based at least in part on the error rate of the test phase.
9. An interconnect device, comprising:
a stream path encoder for encoding a class identifier for a data packet; and
a lane driver for driving the species identifier onto at least one of n data lanes during a non-data time of the data packet.
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14/669,975 | 2015-03-26 | ||
US14/669,975 US9946676B2 (en) | 2015-03-26 | 2015-03-26 | Multichip package link |
CN201680012402.0A CN107430569B (en) | 2015-03-26 | 2016-02-22 | Multi-chip package link |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201680012402.0A Division CN107430569B (en) | 2015-03-26 | 2016-02-22 | Multi-chip package link |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112486875A true CN112486875A (en) | 2021-03-12 |
Family
ID=56975459
Family Applications (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011252637.1A Pending CN112231255A (en) | 2015-03-26 | 2016-02-22 | Multi-chip package link |
CN201680012402.0A Active CN107430569B (en) | 2015-03-26 | 2016-02-22 | Multi-chip package link |
CN202011344684.9A Pending CN112486875A (en) | 2015-03-26 | 2016-02-22 | Multi-chip package link |
Family Applications Before (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011252637.1A Pending CN112231255A (en) | 2015-03-26 | 2016-02-22 | Multi-chip package link |
CN201680012402.0A Active CN107430569B (en) | 2015-03-26 | 2016-02-22 | Multi-chip package link |
Country Status (5)
Country | Link |
---|---|
US (1) | US9946676B2 (en) |
EP (3) | EP3274857B1 (en) |
CN (3) | CN112231255A (en) |
TW (1) | TWI634432B (en) |
WO (1) | WO2016153660A1 (en) |
Families Citing this family (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013105967A1 (en) * | 2012-01-13 | 2013-07-18 | Intel Corporation | Efficient peer-to-peer communication support in soc fabrics |
KR101831550B1 (en) * | 2012-10-22 | 2018-02-22 | 인텔 코포레이션 | Control messaging in multislot link layer flit |
WO2015184994A1 (en) * | 2014-06-04 | 2015-12-10 | Nexpack Limited | Super hub system and method thereof |
US9946676B2 (en) | 2015-03-26 | 2018-04-17 | Intel Corporation | Multichip package link |
DE112015006944B4 (en) * | 2015-09-25 | 2023-03-23 | Intel Corporation | Apparatus, system and method for facilitating communication over a link with a device external to an assembly |
US10709926B2 (en) | 2015-10-06 | 2020-07-14 | Woodway Usa, Inc. | Treadmill |
US10331846B2 (en) * | 2015-12-30 | 2019-06-25 | Arteris, Inc. | Dynamic link serialization in network-on-chip |
US10373657B2 (en) * | 2016-08-10 | 2019-08-06 | Micron Technology, Inc. | Semiconductor layered device with data bus |
US10846258B2 (en) * | 2016-09-30 | 2020-11-24 | Intel Corporation | Voltage modulated control lane |
US11016920B2 (en) * | 2016-12-30 | 2021-05-25 | Intel Corporation | Adaptive calibration technique for cross talk cancellation |
CN109309615B (en) | 2017-07-27 | 2020-07-07 | 华为技术有限公司 | Method and device for transmitting data |
KR102482896B1 (en) | 2017-12-28 | 2022-12-30 | 삼성전자주식회사 | Memory device including heterogeneous volatile memory chips and electronic device including the same |
CN110224846B (en) * | 2018-03-01 | 2022-01-25 | 中兴通讯股份有限公司 | Service protection method based on flexible Ethernet, server and storage medium |
EP3809610B1 (en) * | 2018-07-11 | 2024-03-27 | Huawei Technologies Co., Ltd. | Signal generation device, method, and system |
TWI720345B (en) * | 2018-09-20 | 2021-03-01 | 威盛電子股份有限公司 | Interconnection structure of multi-core system |
US10831535B2 (en) | 2019-01-01 | 2020-11-10 | International Business Machines Corporation | Reducing minimum operating voltage through heterogeneous codes |
CN109857690B (en) * | 2019-01-03 | 2023-04-28 | 华为技术有限公司 | Application system of driver, driver and data transmission method |
CN110162854B (en) * | 2019-05-09 | 2023-05-26 | 重庆大学 | High-speed self-adaptive decision feedback equalizer |
US11815976B2 (en) * | 2019-05-22 | 2023-11-14 | Qualcomm Incorporated | Bandwidth based power management for peripheral component interconnect express devices |
CN114175003B (en) * | 2019-08-07 | 2023-10-10 | 香港大学 | System and method for determining a routing network in a multi-core processor and related multi-core processor |
US11733290B2 (en) * | 2020-03-31 | 2023-08-22 | Advantest Corporation | Flexible sideband support systems and methods |
CN111666235B (en) * | 2020-05-26 | 2022-02-08 | 中国人民解放军国防科技大学 | PIO communication device, chip, equipment and method for high-speed interconnection network interface chip |
US11317242B1 (en) | 2021-01-04 | 2022-04-26 | Bank Of America Corporation | System and method for detecting a distance threshold between a target device and a mobile device has been exceeded |
US20220327074A1 (en) * | 2021-04-13 | 2022-10-13 | SK Hynix Inc. | PERIPHERAL COMPONENT INTERCONNECT EXPRESS (PCIe) SYSTEM AND METHOD OF OPERATING THE SAME |
TWI784500B (en) * | 2021-04-28 | 2022-11-21 | 威鋒電子股份有限公司 | Electronic apparatus and firmware secure update method thereof |
US11592889B2 (en) | 2021-05-12 | 2023-02-28 | Apple Inc. | Die-to-die dynamic clock and power gating |
US20210319841A1 (en) * | 2021-06-24 | 2021-10-14 | Intel Corporation | Systems And Methods For Detecting And Configuring Lanes In A Circuit System |
CN114296430A (en) * | 2021-12-27 | 2022-04-08 | 武汉奥泽电子有限公司 | Multi-state multi-semaphore serial feedback method and system |
Family Cites Families (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1185920A4 (en) | 1998-11-09 | 2005-07-06 | Broadcom Corp | Mixed-signal single-chip integrated system electronics for magnetic hard disk drives |
US7003066B1 (en) | 2001-12-03 | 2006-02-21 | Lattice Semiconductor Corporation | Digital phase locked loop with phase selector having minimized number of phase interpolators |
US7184708B1 (en) | 2003-07-30 | 2007-02-27 | Intel Corporation | Interference mitigation by adjustment of interconnect transmission characteristics |
US7030651B2 (en) * | 2003-12-04 | 2006-04-18 | Viciciv Technology | Programmable structured arrays |
US7512718B2 (en) * | 2005-04-20 | 2009-03-31 | Lawrence J. Dickson | Reconfigurable computing array without chassis |
EP1958039B9 (en) * | 2005-12-02 | 2011-09-07 | Entegris, Inc. | I/o systems, methods and devices for interfacing a pump controller |
US20070239922A1 (en) | 2005-12-09 | 2007-10-11 | Horigan John W | Technique for link reconfiguration |
US7627806B1 (en) | 2006-03-01 | 2009-12-01 | Altera Corporation | Integrated hard-wired or partly hard-wired CRC generation and/or checking architecture for a physical coding sublayer in a programmable logic device |
US7752364B2 (en) | 2006-12-06 | 2010-07-06 | Mosaid Technologies Incorporated | Apparatus and method for communicating with semiconductor devices of a serial interconnection |
US7756123B1 (en) * | 2006-12-21 | 2010-07-13 | Nvidia Corporation | Apparatus, system, and method for swizzling of a PCIe link |
US7991098B2 (en) * | 2007-10-31 | 2011-08-02 | Micron Technology, Inc. | Method and apparatus for training the reference voltage level and data sample timing in a receiver |
US7696782B2 (en) * | 2008-02-15 | 2010-04-13 | Broadcom Corporation | Programmable core for implementing logic change |
US8478554B1 (en) * | 2009-02-09 | 2013-07-02 | Marvell International Ltd. | Reducing eye monitor data samplers in a receiver |
US20100277201A1 (en) * | 2009-05-01 | 2010-11-04 | Curt Wortman | Embedded digital ip strip chip |
US8619605B2 (en) * | 2009-05-13 | 2013-12-31 | Avaya Inc. | Method and apparatus for maintaining port state tables in a forwarding plane of a network element |
US8451969B2 (en) * | 2011-03-15 | 2013-05-28 | Intel Corporation | Apparatus, system, and method for timing recovery |
US8780914B2 (en) | 2011-10-17 | 2014-07-15 | Xilinx, Inc. | Parallel processing of network packets |
US20140003564A1 (en) * | 2012-06-27 | 2014-01-02 | Broadcom Corporation | Multichip synchronization system |
US8686754B2 (en) | 2012-07-05 | 2014-04-01 | Stmicroelectronics International N.V. | Configurable lane architecture in source synchronous systems |
US8832343B2 (en) * | 2012-07-17 | 2014-09-09 | International Business Machines Corporation | Double density I2C system |
US8816717B2 (en) * | 2012-10-17 | 2014-08-26 | International Business Machines Corporation | Reactive material for integrated circuit tamper detection and response |
KR101831550B1 (en) | 2012-10-22 | 2018-02-22 | 인텔 코포레이션 | Control messaging in multislot link layer flit |
US9479456B2 (en) * | 2012-11-02 | 2016-10-25 | Altera Corporation | Programmable logic device with integrated network-on-chip |
US9318408B2 (en) * | 2013-03-12 | 2016-04-19 | Monolithic 3D Inc. | Semiconductor device and structure |
US9153539B2 (en) * | 2013-03-15 | 2015-10-06 | Nvidia Corporation | Ground-referenced single-ended signaling connected graphics processing unit multi-chip module |
CN103257946A (en) * | 2013-05-24 | 2013-08-21 | 浪潮电子信息产业股份有限公司 | High-speed interconnecting method of controllers of tight-coupling multi-control storage system |
US9690741B2 (en) * | 2013-07-15 | 2017-06-27 | Altera Corporation | Configuration via high speed serial link |
US9946676B2 (en) | 2015-03-26 | 2018-04-17 | Intel Corporation | Multichip package link |
-
2015
- 2015-03-26 US US14/669,975 patent/US9946676B2/en active Active
-
2016
- 2016-01-30 TW TW105103036A patent/TWI634432B/en active
- 2016-02-22 CN CN202011252637.1A patent/CN112231255A/en active Pending
- 2016-02-22 EP EP16769240.9A patent/EP3274857B1/en active Active
- 2016-02-22 EP EP20206632.0A patent/EP3846042A1/en active Pending
- 2016-02-22 WO PCT/US2016/018836 patent/WO2016153660A1/en active Application Filing
- 2016-02-22 EP EP20206623.9A patent/EP3796181A1/en active Pending
- 2016-02-22 CN CN201680012402.0A patent/CN107430569B/en active Active
- 2016-02-22 CN CN202011344684.9A patent/CN112486875A/en active Pending
Also Published As
Publication number | Publication date |
---|---|
EP3796181A1 (en) | 2021-03-24 |
WO2016153660A1 (en) | 2016-09-29 |
US20160283429A1 (en) | 2016-09-29 |
CN107430569A (en) | 2017-12-01 |
EP3274857A1 (en) | 2018-01-31 |
EP3274857B1 (en) | 2021-03-24 |
CN107430569B (en) | 2021-06-29 |
TWI634432B (en) | 2018-09-01 |
EP3274857A4 (en) | 2018-11-21 |
EP3846042A1 (en) | 2021-07-07 |
US9946676B2 (en) | 2018-04-17 |
TW201705011A (en) | 2017-02-01 |
CN112231255A (en) | 2021-01-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107430569B (en) | Multi-chip package link | |
US11003610B2 (en) | Multichip package link | |
CN107408032B (en) | Pseudo-random bit sequence in an interconnect | |
CN107015924B (en) | High performance interconnect physical layer | |
US20220350769A1 (en) | Virtualized link states of multiple protocol layer package interconnects | |
JP6745289B2 (en) | Multi-chip package link | |
JP2020201967A (en) | Multichip package link |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |